Expert Adjudication Research Articles

Inter-rater Agreement for Abstraction of the Early Management Bundle, Severe Sepsis/Septic Shock (SEP-1) Quality Measure in a Multi-Hospital Health System

BackgroundThe Early Management Bundle, Severe Sepsis/Septic Shock (SEP-1) quality measure is complex to abstract, which may lead to discrepancies between abstractors. This study was designed to evaluate inter-rater agreement between abstractors at individual hospitals in a health system and a lead abstractor on abstraction elements and measure compliance for SEP-1. MethodsPatient cases qualifying for abstraction for SEP-1 over a four-month period in 2016 were initially abstracted at a local hospital and then centrally by a lead abstractor. Abstraction results were retrospectively compared to determine inter-rater agreement. ResultsA total of 580 SEP-1 cases were abstracted locally and centrally. Each site contributed a median (interquartile range) of 63 (49, 86) cases. There was complete concordance of measure-related elements in 391 cases (67%) (inter-rater agreement: κ = 0.40, p < 0.01). The most common discrepancy (60 cases) was severe sepsis presentation time. There was a weak correlation between SEP-1 compliance adjudicated locally and centrally (r2 = 0.41, p < 0.01). The average change in monthly SEP-1 measure compliance at each site after central adjudication was a 1% increase but ranged from a 49% decrease to a 40% increase. ConclusionsConcordance on SEP-1 abstraction elements between local and expert adjudicators was fair, and SEP-1 performance varied considerably from initial site-reported performance. The detailed nature of SEP-1 can lead to unreliable abstraction, which may lead to inaccurate reporting of compliance with the measure and affect comparability of performance between hospitals. Abstraction by a dedicated team for SEP-1 can reduce variability and improve efficiency.

Development of an Automated Tool for Assessing Response in Patients with Polycythemia Vera

Myeloproliferative neoplasms (MPNs) are a group of hematologic malignancies characterized by the clonal proliferation of one or more hematopoietic cell lineages demonstrated by terminal myeloid cell expansion into the peripheral blood. The most commonly diagnosed Philadelphia chromosome negative (Ph-) MPNs include essential thrombocythemia (ET), polycythemia vera (PV), and myelofibrosis (MF).Assessing various categories of response to a treatment is determined according to multiple factors, including laboratory values; transfusion dependence; degree of splenomegaly; mutational status and variant allelic frequency (VAF); cytogenetic abnormalities; and bone marrow morphological features, such as myeloblasts, fibrosis, and the volume ratio of hematopoietic stem cells in bone marrow [1]. While many of these components exist in electronic health records (EHRs), they are documented with varying degrees of structure. Some terms are recorded as tabular values, whereas others are captured purely in free text (Table 1). However, nearly all clinical trials and reviews studying patients with MPNs rely on expert adjudication of response as determined by these data points.The current process for determining response involves manual review of the patient's EHR data, an arduous task requiring extensive human effort. To alleviate these efforts, the Richard T. Silver Myeloproliferative Neoplasm Center at Weill Cornell Medicine (WCM) worked in tandem with the Architecture for Research Computing in Health (ARCH) program to develop a method for assessing response in patients with MPNs. A research data repository (RDR) containing data from both outpatient and inpatient EHRs was designed to allow for computational assessment of response.Structured data elements, including laboratory values, mutational data, and VAF, were extracted from the EHR. A natural language processing (NLP) pipeline using the Leo framework was developed [2] to extract data on cellularity, reticulin fibrosis, and myeloblast count from bone marrow biopsy pathology reports (Figure 1). Other data points, including splenomegaly and symptom burden, continue to require interpretation and manual collection by trained research personnel, who enter these data points into REDCap (Research Electronic Data Capture), a WCM-provisioned secure web application for managing clinical databases. These manual data are then pivoted and loaded into a Microsoft SQL Server environment [3]. Data generated by the MPN RDR were frequently reviewed for quality control, which drove subsequent iterations designed to minimize any identified errors.After acceptable confidence levels had been achieved, the MPN RDR was queried to provide data that were used to contribute to response assessments [1] for a retrospective review studying PV patients. Fully automated response parameters included hematocrit, platelet, and white blood cell count values; cellularity; reticulin fibrosis; and JAK2V617F VAF. Partially automated response assessments included rates of phlebotomy. Manually collected response criteria included symptom burden, degree of palpable splenomegaly, and indications of hemorrhagic/thrombotic events. Extracted data were merged with manually collected data within clinically justified temporal windows and applied to PV response criteria. The process provided valuable insight on potential modifications to the extraction process.Future steps include extending these processes to criteria still dependent on manual collection. Efforts are currently under way to apply NLP to hepatosplenomegaly and cytogenetics reports. This process will also be used to assist in response assessments for the spectrum of other MPN subtypes. While the feasibility of a fully comprehensive computable approach to assessing response in MPNs may not be entirely feasible nor even advisable, it is rather the objective that clinical data collection, which has been historically onerous, become automated to the furthest extent possible so as to allow research personnel to focus on the extraction of data elements that do require manual adjudication. Adaption of a similar workflow may help other institutions expand their ability to assess response in patients with MPNs and, potentially, additional hematologic malignancies. DisclosuresNo relevant conflicts of interest to declare.

1948. A Host-Response Assay Distinguishes Between Simple Influenza Patients and Influenza Patients With Bacterial Coinfection

BackgroundIdentifying bacterial coinfection in influenza patients can be difficult as the symptoms of simple influenza vs. mixed infections are often similar, leading to antibiotic overuse. A new host-response assay (ImmunoXpert™) that integrates the levels of three proteins (TRAIL, IP-10, and CRP) was shown to exhibit high performance in distinguishing between bacterial and viral disease in two double-blind validation studies. Here we sought to evaluate its ability to differentiate between simple influenza and influenza with bacterial coinfection.MethodsThe study population included 653 febrile pediatric and adult patients prospectively recruited in the “Curiosity” study. Patient etiology (simple viral vs. mixed infection) was determined by unanimous expert adjudication based on comprehensive clinical, laboratory and radiological assessment. Influenza strains (A or B) were detected using multiplex PCR applied to nasal swabs (Seeplex-RV15). We compared the expert panel diagnosis with the assay that gives three possible outcomes: viral, bacterial (including viral with bacterial coinfection) or equivocal. An equivocal outcome does not provide diagnostic information and is observed in ~10% of cases.ResultsOut of 653 patients, 51 had positive influenza detection and unanimous expert diagnosis: 44 simple viral infections and seven influenza with bacterial coinfections (Figure 1). Antibiotics were prescribed to all seven cases of influenza with bacterial coinfection and to 20/44 cases adjudicated as simple viral infections, indicating an overuse rate of 45%. The assay correctly classified 40 of the 44 simple viral cases (out of the remaining four, two were assigned viral with bacterial coinfection, and two received equivocal outcomes) as well as five of the seven viral with bacterial coinfection cases (the remaining two received equivocal outcomes) supporting the assay’s potential to reduce antibiotic overuse 5-fold (from 45% to 4/44 = 9%, P < 0.001).ConclusionThe host–response assay can differentiate between simple influenza and influenza patients with bacterial coinfection, with potential to reduce antibiotic overuse. Utility studies are warranted to demonstrate that the assay can safely assist physicians in correct management of influenza patients. Disclosures M. Paz, MeMed Diagnostics: Employee, Salary. K. Oved, MeMed Diagnostics: Board Member, Employee and Shareholder, Salary. T. Gottlieb, MeMed Diagnostics: Employee, Salary. A. Cohen, MeMed Diagnostics: Employee, Salary. R. Navon, MeMed Diagnostics: Employee, Salary. N. Mastboim, MeMed Diagnostics: Employee, Salary. E. Bamberger, MeMed Diagnostics: Employee, Salary. T. Friedman, MeMed Diagnostics: Employee, Salary. L. Etshtein, MeMed Diagnostics: Employee, Salary. O. Boico, MeMed Diagnostics: Employee, Salary. I. Potasman, MeMed Diagnostics: Holding stock options, Stock options. E. Eden, MeMed Diagnostics: Board Member, Employee and Shareholder, Salary. L. Shani, MeMed Diagnostics: Employee, Salary.

THUR 121 Identifying participants with parkinson’s disease in uk biobank

BackgroundLinkage to routinely collected NHS data from primary, secondary care and death certificates enables identification of participants with Parkinson’s Disease (PD) within the UK Biobank cohort of 5 00 000 adults. Validation of the accuracy of this data is required prior to their use in research studies.MethodIn this validation study participants (n=125) with a code indicating PD were identified from a sample of 17 000 participants in the cohort. Diagnoses were validated by expert adjudicators, based on free text electronic medical records. Positive predictive values (PPV,% of cases identified that are true cases) were calculated.ResultsPrimary care diagnostic codes identified 93% of PD cases, with a PPV of 95%. Combined secondary care and death data identified 42% of PD cases with a PPV of 84%.Combining diagnostic and medication codes identified more participants, but did not increase the PPV.ConclusionsThis study suggests that linkage to routinely collected healthcare data is a reliable method for identifying participants with PD in the UK Biobank cohort.Primary care diagnostic codes identified the highest proportion of participants and had the highest PPV, demonstrating the value of using primary care data to identify cases of disease in large population based cohort studies.

P1625Identification of chronic thromboembolic pulmonary hypertension on standard computed tomography pulmonary angiography for suspected acute pulmonary embolism

Background: Chronic thromboembolic pulmonary hypertension (CTEPH) is often diagnosed after a long delay, even though signs of CTEPH may already be present on the computed tomography pulmonary angiogram (CTPA) used to diagnose a preceding acute pulmonary embolism (PE). Diagnostic delay may be associated with worse prognosis. In this study we evaluated the diagnostic accuracy of extensive CTPA reading for predicting a future CTEPH diagnosis after acute PE. Method(s): Three expert radiologists scored radiological signs of CTEPH on CTPA performed for suspected acute PE in 50 PE patients who were subsequently diagnosed with CTEPH during follow-up with right heart catheterisation (cases), and in 50 PE patients, in whom sequential echocardiograms performed >2 years after the acute PE diagnosis had not shown any signs of pulmonary hypertension (controls). Cases and controls were matched on RV-to-LV ratio on baseline CTPA. Sensitivity and specificity of expert CTPA reading for predicting future CTEPH was calculated, and best predicting radiological parameters were identified by conditional backward multivariate analysis. Result(s): 75% of cases had signs of acute PE and 94% signs of chronic PE and/or PH. The overall expert adjudication yielded a sensitivity of 72% (95% CI 58-84%) and a specificity of 94% (95% CI 83-99%) for a future CTEPH diagnosis. We identified six objective, independent radiological predictors of CTEPH: intravascular webs, pulmonary artery retraction or dilatation, bronchial artery dilatation, right ventricle hypertrophy and interventricular septum flattening (Figure 1). The presence of >=3 of these parameters was associated with a sensitivity of 70% (95% CI 55-82%), a specificity of 96% (95% CI 86-100%) and a c-statistic of 0.92. Conclusion(s): In our cohort, expert radiologists identified 36 out of 50 PE patients with a future CTEPH diagnosis. The most probable explanation is that most of them already had CTEPH at the moment of the PE diagnosis. Our findings support the hypothesis that dedicated CTPA reading in patients with confirmed PE may lead to faster CTEPH diagnosis.

Lessons Learned in the Development of a Computable Phenotype for Response in Myeloproliferative Neoplasms.

Determining response status in patients with myeloproliferative neoplasms is a complex problem requiring the integration of both structured and unstructured data elements from disparate information systems. By applying multiple techniques, a collaborative team of informatics professionals and research personnel were able to determine which elements were amenable to automated extraction and which required expert adjudication. With this knowledge in mind, we were able to build a system that joins together programmatically-derived and manually-abstracted data elements to facilitate response assessment - an important end point in clinical and translational research in this disease area.

472 - NUDT15 Variants Contribute to Thiopurine-Induced Myelosuppression in European Populations

Introduction Thiopurines are commonly used in the maintenance treatment of inflammatory bowel disease (IBD) but this is limited by myelosuppression in 7% of patients.1 Thiopurine S-methyltransferase (TPMT) variants only explain 20% of thiopurine-induced myelosuppression (TIM) in Europeans suggesting the presence of other genetic determinants.2 We aimed to identify the clinical features of TIM and to identify novel variants associated with TIM through a genome wide association study and whole exome sequencing. Method We recruited 491 IBD patients with TIM from 82 hospitals. Inclusion criteria included drug exposure in the 7 days prior to TIM, white blood cell count ≤2.5 ×109/L, or neutrophil count ≤1.0 ×109/L and that TIM necessitated dose reduction/drug withdrawal. After expert adjudication, 329 cases assigned a high likelihood of causality were genotyped and exome sequenced with 635 thiopurine-tolerant IBD controls for use in the final analyses. Results We first confirmed an association of TIM with TPMT in an initial GWAS. We then, using exome sequencing, discovered a novel 6 bp in-frame deletion (rs746071566; AGGAGTC/A,p.Gly17_Val18del) at position 48611918 of chromosome 13 in exon 1 of NUDT15 (5.8% cases, 0.2% controls, OR=38.2,p=1.33×10-8). This deletion was replicated in an independent cohort (3/75 [4%] cases vs. 2/785 [0.3%] thiopurine-tolerant IBD controls; OR=16.2, p=0.006). All NUDT15 coding variants were confirmed with complete concordance by Sanger sequencing. Multivariable logistic regression demonstrated that adjusted dose (OR 2.2,p=9.4×10-11), NUDT15 genotype (OR 21.7,p=2.0×10-8) and TPMT genotype (OR 2.2,p=2.6×10-4 for MUT/WT and OR 51.2,p=1.8×10-4 for MUT/MUT) were independently associated with TIM. Conclusion These are the largest ever clinical and genetic analyses of thiopurine-induced myelosuppression. We identified NUDT15 coding variants, including a novel 6 bp deletion; carriage of any coding variant confers a 22-fold increase in the odds of TIM, independent of TMPT genotype and thiopurine dose. Although NUDT15 variants are less common than TPMT variants, their effect size for heterozygotes is greater. The number of patients needed to genotype to prevent TIM due to NUDT15 coding variant heterozygosity is 100. The estimated absolute risk of TIM in NUDT15 heterozygotes is 59%. A future clinical decision tool incorporating NUDT15 and TPMT genotypes will offer personalised thiopurine therapy and prevent the considerable morbidity and costs associated with severe bone marrow toxicity. Reference . Gisbert JP & Gomollon F.doi10.1111/j.1572-0241.2008.01848.x (2008) . Yang SK et al. doi10.1038/ng.3060 (2014) Disclosure of Interest None Declared

A host-protein signature is superior to other biomarkers for differentiating between bacterial and viral disease in patients with respiratory infection and fever without source: a prospective observational study

Bacterial and viral infections often present with similar symptoms. Etiologic misdiagnosis can alter the trajectory of patient care, including antibiotic overuse. A host-protein signature comprising tumor necrosis factor-related apoptosis-inducing ligand (TRAIL), interferon gamma-induced protein-10 (IP-10), and C-reactive protein (CRP) was validated recently for differentiating bacterial from viral disease. However, a focused head-to-head comparison of its diagnostic performance against other biomarker candidates for this indication was lacking in patients with respiratory infection and fever without source. We compared the signature to other biomarkers and prediction rules using specimens collected prospectively at two secondary medical centers from children and adults. Inclusion criteria included fever > 37.5 °C, symptom duration ≤ 12 days, and presentation with respiratory infection or fever without source. Comparator method was based on expert panel adjudication. Signature and biomarker cutoffs and prediction rules were predefined. Of 493 potentially eligible patients, 314 were assigned unanimous expert panel diagnosis and also had sufficient specimen volume. The resulting cohort comprised 175 (56%) viral and 139 (44%) bacterial infections. Signature sensitivity 93.5% (95% CI 89.1–97.9%), specificity 94.3% (95% CI 90.7–98.0%), or both were significantly higher (all p values < 0.01) than for CRP, procalcitonin, interleukin-6, human neutrophil lipocalin, white blood cell count, absolute neutrophil count, and prediction rules. Signature identified as viral 50/57 viral patients prescribed antibiotics, suggesting potential to reduce antibiotic overuse by 88%. The host-protein signature demonstrated superior diagnostic performance in differentiating viral from bacterial respiratory infections and fever without source. Future utility studies are warranted to validate potential to reduce antibiotic overuse.

658: Prediction of pregnancy outcomes from serum and clinical factors in women receiving 17-alpha hydroxyprogesterone caproate (17-OHPC)

We sought to improve prediction of spontaneous preterm birth (SPTB) and delivery gestational age (GA) in women with a prior SPTB receiving 17-OHPC by combining clinical factors with novel serum biomarkers. This is a secondary analysis of a prospective observational study of 5,501 women with singleton gestations at 11 US sites with the aim of developing SPTB biomarkers. For this analysis, we included women with a prior SPTB who received 17-OHPC; this subset was excluded from the parent analysis a priori. All preterm birth (PTB) cases were subject to expert adjudication masked to proteomic results. Medically indicated PTB were excluded. Targeted mass spectrometry quantified peptides from 63 proteins in serum drawn at 170/7 - 246/7 weeks. Clinical and proteomic profiles were examined to identify predictors of delivery GA (Outcome #1) and SPTB (Outcome #2). We evaluated models for correlation with delivery GA and AUC for SPTB prediction. Regularized regression selected significant predictors and Kaplan-Meier analysis estimated survival curves. ANOVA was used to compare models. Unsupervised hierarchical clustering was used to explore biological pathways of predictive peptides. 80 women met inclusion criteria. Serum was collected at a median of 4 (IQR 3 - 6) weeks after 17-OHPC initiation. Delivery was at a median of 37.6 (IQR 35 - 39) weeks; 42.5% had recurrent SPTB. In clinical-only models, education (<high school; mean effect -18 days) and cervical length (<25mm; mean effect -16 days) were associated with outcomes (Table). In peptide-only models, complement factor B (CFAB) and inhibin beta C chain (INHBC) peptides were the best predictors of delivery GA (Outcome #1). Each 2-fold increase in peptide levels prolonged GA by 6 (CFAB) and 26 (INHBC) days. The clinical + peptide models improved delivery GA and SPTB prediction compared to clinical-only models (Figure). CFAB alone also improved recurrent SPTB prediction over clinical-only models (Outcome #2). CFAB and INHBC clustered in a group of 10 inflammatory/stress proteins. Adding serum peptides to clinical risk factors improves prediction of delivery GA and recurrent SPTB in high-risk women receiving 17-OHPC. Outcomes in these women may be related to inflammatory mediators measurable in serum in mid-trimester prior to onset of clinical symptoms.View Large Image Figure ViewerDownload Hi-res image Download (PPT)

A Novel Host-protein Assay Accurately Distinguishes Bacterial From Viral Upper Respiratory Tract Infections

BackgroundBacterial and viral infections are often clinically indistinguishable, particularly in upper respiratory tract infections (URTI), which leads to antibiotic misuse. A novel assay (ImmunoXpert™) that integrates measurements of three host-response proteins (TRAIL, IP-10, CRP) was recently developed to assist in differentiation between bacterial and viral etiologies. We evaluated the assay performance in URTI patients and compared it with standard laboratory measures.MethodsWe performed a sub-analysis of 464 patients with clinical suspicion of URTI enrolled in three previously conducted multi-center clinical studies that evaluated the assay performance in patients with acute infections: ‘Curiosity’ study (NCT01917461), ‘Opportunity’ study (NCT01931254), and ‘Pathfinder’ study (NCT01911143). Comparator method was predetermined criteria combined with expert panel adjudication, which was blinded to the test results. Diagnostic performance was evaluated by comparing test and comparator method outcomes.ResultsA unanimous panel adjudication was attained for 61 bacterial (13%) and 241 viral (52%) patients (162 patients (35%) had an indeterminate diagnosis). The assay distinguished between bacterial and viral infected patients with a sensitivity of 92% (95% CI: 82%- 98%) and specificity of 93% (88%-96%) with 11% equivocal test results. Overall the assay outperformed other routine laboratory tests (FIG 1), including: white blood cell count (WBC; cutoff 15,000 cells/µL, sensitivity 48% (35%-60%), P < 10-−6; specificity 85% (80%-90%), P < 0.05); CRP (cutoff 40 mg/L, sensitivity 82% (72%–92%), P = 0.16, specificity 79% (74%–84%), P < 10-4); Procalcitonin (PCT; cutoff 0.5 ng/mL, sensitivity 22% (11%–32%), P < 10–14, specificity 80% (74%–85%), P < 0.001); absolute neutrophil count (ANC; cutoff 10,000 cells/µL, sensitivity 58% (45%–71%), P < 10-−4, specificity 94% (91%–97%), P = 0.7).ConclusionThe novel assay demonstrated superior performance compared with routine laboratory tests (WBC, ANC) and biomarkers (CRP, PCT), in distinguishing bacterial from viral etiologies in patients with URTI. It has the potential to help clinicians avoid missing bacterial infections or prescribing unwarranted antibiotics for viral URTIs.Disclosures K. Oved, MeMed Diagnostics: Board Member, Employee and Shareholder, Salary E. Eden, MeMed Diagnostics: Board Member, Employee and Shareholder, Salary T. Gottlieb, MeMed Diagnostics: Employee, Salary R. Navon, MeMed Diagnostics: Employee, Salary A. Cohen, MeMed Diagnostics: Employee, Salary O. Boico, MeMed Diagnostics: Employee, Salary M. Paz, MeMed Diagnostics: Employee, Salary L. Etshtein, MeMed Diagnostics: Employee, Salary G. Kronenfeld, MeMed Diagnostics: Employee, Salary T. Friedman, MeMed Diagnostics: Employee, Salary E. Bamberger, MeMed Diagnostics: Employee, Salary I. Chistyakov, MeMed Diagnostics: Consultant, Consulting fee I. Potasman, MeMed Diagnostics: Holding stock options, stock options

Estimating the prevalence of generalized and partial lipodystrophy: findings and challenges.

BackgroundLipodystrophy (LD; non-human immunodeficiency virus [HIV]-associated) syndromes are a rare body of disorders for which true prevalence is unknown. Prevalence estimates of rare diseases are important to increase awareness and financial resources. Current qualitative and quantitative estimates of LD prevalence range from ~0.1 to 90 cases/million. We demonstrate an approach to quantitatively estimate LD prevalence (all, generalized, and partial) through a search of 5 electronic medical record (EMR) databases and 4 literature searches.MethodsEMR and literature searches were conducted from 2012 to 2014. For the EMR database searches (Quintiles, IMS LifeLink, General Electric Healthcare, and Humedica EMR), LD cases were identified by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 272.6 (United Kingdom General Practice Research Database used other diagnostic codes to identify LD) plus additional LD-associated clinical characteristics (patients with HIV or documented HIV treatment were excluded). Expert adjudication of cases was used for the Quintiles database only. Literature searches (PubMed and EMBASE) were conducted for each of the 4 major LD subtypes. Prevalence estimates were determined by extrapolating the total number of cases identified for each search to the database population (EMR search) and European population (literature search).ResultsThe prevalence range of all LD across all EMR databases was 1.3–4.7 cases/million. For the adjudicated Quintiles search, the estimated prevalence of diagnosed LD was 3.07 cases/million (95% confidence interval [CI], 2.30–4.02), 0.23 cases/million (95% CI, 0.06–0.59) and 2.84 cases/million (95% CI, 2.10–3.75) for generalized lipodystrophy (GL) and partial lipodystrophy (PL), respectively. For all literature searches, the prevalence of all LD in Europe was 2.63 cases/million (0.96 and 1.67 cases/million for GL and PL, respectively).ConclusionLD prevalence estimates are at the lower range of previously established numbers, confirming that LD is an ultra-rare disease. The establishment of diagnostic criteria and coding specific to the 4 major LD subtypes and future studies/patient registries are needed to further refine our estimates.

IDENTIFYING DEMENTIA CASES IN PROSPECTIVE COHORT STUDIES USING ROUTINELY COLLECTED HEALTH DATASETS

Prospective population-based cohort studies that recruit participants in mid-life are potentially rich resources for studying dementia aetiology. In many cohorts, disease cases are identified through linkage to routinely-collected health datasets (RCHDs) such as hospital admissions, deaths and primary care data. We investigated the accuracy of these datasets to identify dementia cases. We conducted a systematic review of published validation studies. We searched five electronic databases for studies comparing coding of dementia in RCHDs to an expert-led reference standard and reporting either positive predictive value (PPV, proportion of true positives among identified cases) or sensitivity (proportion of identified cases among true positives). We then performed our own validation study of dementia coding in RCHDs using data from UK Biobank (UKB). UKB is a large, population-based prospective cohort study of >500,000 participants aged 40–69 years when recruited in 2006–2010. Having created a list of ICD-10 and primary care (Read V2) dementia codes, we identified Edinburgh-based UKB participants with ≥1 dementia code in hospital admissions, deaths or primary care data. Using extracted full text correspondence from the electronic medical record (EMR), a neurologist adjudicated on whether dementia was present, providing the reference standard to which the coded data were compared. We calculated the PPV for each dataset. We identified 22 eligible studies in the systematic review, providing 20 estimates of PPV and eight of sensitivity. Study methods varied widely. For all-cause dementia, PPV estimates ranged from 33–100%, but many were >80% (fig1). PPVs for Alzheimer's disease were higher than those for vascular dementia (fig2). Sensitivity estimates ranged from 21–86%, depending on the data sources. In our UKB validation study, among 17,000 Edinburgh-based participants, 125 had ≥1 dementia code. Expert adjudication in an initial subset of 44 participants demonstrated PPVs of 41/44 (93%, 95% CI 81–99) overall, 13/15 (87%, 95% CI 60–98) for hospital admissions, 2/2 (100%, 95% CI 16–100) for deaths and 33/34 (97%, 95% CI 85–100) for primary care data.

Preliminary Results of a Novel Algorithmic Method Aiming to Support Initial Causality Assessment of Routine Pharmacovigilance Case Reports for Medication-Induced Liver Injury: The PV-RUCAM.

Data incompleteness in pharmacovigilance (PV) health records limits the use of current causality assessment methods for drug-induced liver injury (DILI). In addition to the inherent complexity of this adverse event, identifying cases of high causal probability is difficult. The aim was to evaluate the performance of an improved, algorithmic and standardised method called the Pharmacovigilance-Roussel Uclaf Causality Assessment Method (PV-RUCAM), to support assessment of suspected DILI. Performance was compared in different settingswith regard to applicability and differentiation capacity. A PV-RUCAM score was developed based on the seven sections contained in the original RUCAM. The score provides cut-off values for or against DILI causality, and was applied on two datasets of bona fide individual case safety reports (ICSRs) extracted randomly from clinical trial reports and a third dataset of electronic health records from a global PV database. The performance of PV-RUCAM adjudication was compared against two standards: a validated causality assessment method (original RUCAM) and global introspection. The findings showed moderate agreement against standards. The overall error margin of no false negatives was satisfactory, with 100% sensitivity, 91% specificity, a 25% positive predictive value and a 100% negative predictive value. The Spearman's rank correlation coefficient illustrated a statistically significant monotonic association between expert adjudication and PV-RUCAM outputs (R=0.93). Finally, there was high inter-rater agreement (K w=0.79) between two PV-RUCAM assessors. Within the PV setting of a pharmaceutical company, the PV-RUCAM has the potential to facilitate and improve the assessment done by non-expert PV professionals compared with other methods when incomplete reports must be evaluated for suspected DILI. Prospective validation of the algorithmic toolis necessary prior to implementation for routine use.

Hyperglycemia predicts poststroke infections in acute ischemic stroke

To investigate whether admission hyperglycemia predicts poststroke infections and, if so, whether poststroke infections modify the effect of admission hyperglycemia on functional outcome in ischemic stroke. We used data from acute ischemic stroke patients in the Preventive Antibiotics in Stroke Study (PASS), a multicenter randomized controlled trial (n = 2,550) investigating the effect of preventive antibiotics on functional outcome. Admission hyperglycemia was defined as blood glucose ≥7.8 mmol/L and poststroke infection as any infection during admission judged by an expert adjudication committee. Functional outcome at 3 months was assessed with the modified Rankin Scale. Of 1,676 nondiabetic ischemic stroke patients, 338 (20%) had admission hyperglycemia. After adjustment for potential confounding variables, admission hyperglycemia was associated with poststroke infection (adjusted odds ratio [aOR] 2.31, 95% CI 1.31-4.07), worse 3-month functional outcome (common aOR 1.40, 95% CI 1.12-1.73), and 3-month mortality (aOR 2.11, 95% CI 1.40-3.19). Additional adjustment for poststroke infection in the functional outcome analysis, done to assess poststroke infection as an intermediate in the pathway from admission hyperglycemia to functional outcome, did not substantially change the model. In patients with recorded diabetes mellitus (n = 418), admission hyperglycemia was not associated with poststroke infection (aOR 0.49, 95% CI 0.15-1.58). In nondiabetic acute ischemic stroke patients, admission hyperglycemia is associated with poststroke infection and worse functional outcome. Poststroke infections did not modify the effect of admission hyperglycemia on functional outcome in ischemic stroke.

Evaluating appropriate red blood cell transfusions: a quality audit at 10 Ontario hospitals to determine the optimal measure for assessing appropriateness.

Evaluating the appropriateness of red blood cell (RBC) transfusion requires labor-intensive medical chart audits and expert adjudication. We sought to determine the appropriateness of RBC transfusions at 10 hospitals using retrospective chart review and to determine whether simple metrics (proportion of single-unit transfusions, RBCs/100 acute inpatient days, proportion of transfusions with pretransfusion hemoglobin <80 g/L or posttransfusion hemoglobin <90 g/L) could be used as surrogate markers of appropriateness by comparing their values with the results from the audit. An initial block of 30 RBC units was dually adjudicated for appropriateness followed by additional blocks of 10 units until the difference between the cumulative percentage of appropriate RBC units in the preceding block and final block was <3%. Pearson correlation tests were used to evaluate associations between the metrics and percentages of appropriate transfusions per hospital. Two-by-two tables were used to assess the utility of the metrics to classify transfusions for appropriateness. Of the 498 units audited, 78% were adjudicated as appropriate (κ = 0.9603), with significant variability between institutions (p < 0.0001). Fifty audits or less were required at nine of the institutions. The values of the metrics were not found to have significant correlations with appropriateness, and the metric that misclassified the smallest proportion of transfusions for appropriateness was pretransfusion hemoglobin <80 g/L, at 24%. Our findings suggest that a chart audit of 50 RBC transfusions with adjudication using robust criteria is the optimal means of evaluating RBC transfusion appropriateness at an institution for benchmarking and quality-improvement initiatives.

Incidence of transfusion reactions: a multicenter study utilizing systematic active surveillance and expert adjudication.

Prevalence estimates of the serious hazards of transfusion vary widely. We hypothesized that the current reporting infrastructure in the United States fails to capture many transfusion reactions and undertook a multicenter study using active surveillance, data review, and adjudication to test this hypothesis. A retrospective record review was completed for a random sample of 17% of all inpatient transfusion episodes over 6 months at four academic tertiary care hospitals, with an episode defined as all blood products released to a patient in 6 hours. Data were recorded by trained clinical research nurses, and serious reactions were adjudicated by a panel of transfusion medicine experts. Of 4857 transfusion episodes investigated, 1.1% were associated with a serious reaction. Transfusion-associated circulatory overload was the most frequent serious reaction noted, being identified in 1% of transfusion episodes. Despite clinical notes describing a potential transfusion association in 59% of these cases, only 5.1% were reported to the transfusion service. Suspected transfusion-related acute lung injury/possible transfusion-related acute lung injury, anaphylactic, and hypotensive reactions were noted in 0.08, 0.02, and 0.02% of transfusion episodes, respectively. Minor reactions, including febrile nonhemolytic and allergic, were noted in 0.62 and 0.29% of transfusion episodes, respectively, with 30 and 50% reported to the transfusion service. Underreporting of cardiopulmonary transfusion reactions is striking among academic, tertiary care hospitals. Complete and accurate reporting is essential to identify, define, establish pathogenesis, and mitigate/treat transfusion reactions. A better understanding of the failure to report may improve the accuracy of passive reporting systems.

No Association between Dietary Patterns and Risk for Cognitive Decline in Older Women with 9-Year Follow-Up: Data from the Women’s Health Initiative Memory Study

BackgroundData on the association between dietary patterns and age-related cognitive decline are inconsistent. ObjectiveTo determine whether dietary patterns assessed by the alternate Mediterranean diet score (aMED), the Healthy Eating Index-2010 (HEI-2010), the Alternate Healthy Eating Index 2010 (AHEI-2010), or the Dietary Approach to Stop Hypertension (DASH) diet score are associated with cognitive decline in older women, and to examine whether dietary patterns modify the risk for cognitive decline in women with hypertension. DesignProspective, longitudinal cohort study. Food frequency questionnaires were used to derive dietary patterns at baseline. Hypertension was defined as self-report of current drug therapy for hypertension or clinic measurement of systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg. Participants and settingPostmenopausal women (N=6,425) aged 65 to 79 years who participated in the Women’s Health Initiative Memory Study and were cognitively intact at baseline. Main outcome measuresCognitive decline was defined as cases of mild cognitive impairment (MCI) or probable dementia (PD). Cases were identified through rigorous screening and expert adjudication. Statistical analyses performedCox proportional hazards models with multivariable adjustment were used to estimate the relative risk for developing MCI or PD. ResultsDuring a median follow-up of 9.11 years, we documented 499 cases of MCI and 390 of PD. In multivariable analyses we did not detect any statistically significant relationships across quintiles of aMED, HEI-2010, DASH, and AHEI-2010 scores and MCI or PD (P values for trend=0.30, 0.44, 0.23, and 0.45). In women with hypertension, we found no significant association between dietary patterns and cognitive decline (P values for trend=0.19, 0.08, 0.07, and 0.60). ConclusionsDietary patterns characterized by the aMED, HEI-2010, AHEI-2010, or DASH dietary score were not associated with cognitive decline in older women. Adherence to a healthy dietary pattern did not modify the risk for cognitive decline in women with hypertension.

Piper Study Rationale and Design: Characterizing the Incidence of Acute Respiratory Distress Syndrome in Hematology-Oncology Patients Following the Transfusion of Conventional or Pathogen-Reduced Platelet Components

The INTERCEPT™ Blood System (amotosalen + UVA illumination) for platelets is FDA-approved in the USA for the ex vivo preparation of pathogen-reduced, apheresis platelet components (PCs) to reduce transfusion-transmitted infection, including sepsis, and to potentially reduce the risk of transfusion-associated graft vs. host disease. Platelet transfusion is a potential risk factor for acute respiratory distress syndrome (ARDS), but the incidence of ARDS following platelet transfusion is not well-defined. A phase 3 study (SPRINT) of patients treated with INTERCEPT PCs vs. conventional PCs reported a frequency of ARDS of 1.6% vs. 0%, respectively (Snyder et al., Transfusion 2005;45:1864-1875). This unexpectedly low incidence of ARDS, owing to the lack of precise definitions, prompted the conduct of a blinded reanalysis using the American European Consensus Criteria for ARDS. Reanalysis demonstrated a frequency of ARDS of 6.0% and 4.9% for recipients of pathogen-reduced PCs and conventional PCs, respectively, with no significant treatment difference between groups (Corash et al., Blood 2011;117:1014-1020). Large scale active hemovigilance programs in Europe conducted by regulatory authorities from 2009-2014 reported no differences in the incidence of ARDS following platelet transfusion (0.33 vs. 0.32 cases per 10,000 platelets transfused). To evaluate the incidence of severe pulmonary adverse events requiring assisted ventilation in hematology-oncology (heme/onc) patients following the transfusion of apheresis derived INTERCEPT PCs compared to transfusion of conventional apheresis PCs. A prospective, non-randomized sequential cohort, open label, multi-center, non-inferiority, phase 4 surveillance study following transfusion of INTERCEPT PCs will be conducted. The patient population will be heme/onc hematopoietic stem cell transplant (HSCT) patients expected to require >1 PC transfusions. The proportion of patients requiring treatment-emergent assisted mechanical ventilation (TEAMV) during the observation period as defined by intubation or tight-fitting mask. TEAMV provides an objective measure of severe, clinically significant pulmonary injury. Patients will be assessed for TEAMV within 7 days after each study PC transfusion. 3000 patients allow for >90% power (one-sided alpha 0.025) for a non-inferiority margin of 0.023, assuming an event rate of 3.7% for the Control group. Events will be adjudicated by a blinded Pulmonary Expert Adjudication Panel. The PIPER study will address a significant deficit in the understanding of serious pulmonary events in heme/onc HSCT patients who require platelet support. It will be the first large controlled trial designed and powered to detect serious pulmonary adverse events in this patient population.

Identifying the Impact of Palliative Care Interventions: A Systematic Review (SA538D)

•Identify the current evidence for palliative care interventions to improve patient and caregiver outcomes in seriously ill populations.•Identify common methodological limitations among randomized controlled trials of palliative care interventions.•Identify priority areas for future research in palliative care. A 2008 systematic review described sparse evidence in support of palliative care; however, several high-impact trials have been conducted since then. Increasing calls for adoption of palliative care in serious illness warrant systematic re-examination of the evidence base. We sought to characterize the structure, processes, outcomes, and quality of randomized controlled trials (RCTs) of palliative care interventions in serious illness. With the aid of a research librarian, we searched the Medline, EMBASE, CINAHL, and Cochrane databases from inception through February, 2015. We included RCTs of interventions comprised of ≥2 palliative care domains (as outlined by the National Consensus Project for Quality Palliative Care) and conducted in seriously ill adult populations. Eligible studies reported on ≥1 of the following predefined outcomes: patient quality of life (QOL), patient symptom burden (physical and psychological), satisfaction with care, advance care planning, mortality, healthcare utilization, costs, caregiver QOL, and caregiver burden. Two investigators independently screened, reviewed, extracted data from, and assessed the quality of each record using standardized criteria; disagreements were resolved by discussion or expert adjudication. We identified 6,143 records; 54 (0.009%) articles met all inclusion criteria, representing 41 trials. Substantial heterogeneity exists in the design, content, and outcomes of palliative care interventions, limiting the ability to pool results. The greatest evidence for the benefit of palliative care was found regarding: satisfaction with care (78% of studies favoring palliative care), patient QOL (63%), and caregiver burden (63%). Interventions were less often associated with benefit regarding: symptoms (35%), mortality (25%), and caregiver QOL (20%). Common methodological limitations included inadequate description of intervention content; poor recruitment and follow-up; inconsistent use of outcome measures, many of which were not designed for seriously ill populations; and, flaws in intent-to-treat analyses. We will discuss priorities for future research, including the need to define a common core of high-priority outcome measures and to investigate “active ingredients” of interventions.

Plasma phospholipids and prevalence of mild cognitive impairment and/or dementia in the ARIC Neurocognitive Study (ARIC-NCS)

IntroductionPhospholipids are altered in brains of patients with dementia and some studies suggest their plasma levels may be useful in the detection of mild cognitive impairment (MCI) and dementia. MethodsWe measured 188 plasma metabolites in participants who underwent a detailed neuropsychological assessment and classified as normal (n = 153), MCI (n = 145), or dementia (n = 143) by expert adjudication. ResultsAmong 10 phospholipids recently implicated as altered in dementia, higher concentration of PC aa C36:6 was significantly associated with decreased prevalence of dementia (odds ratio = 0.71, 95% confidence interval = 0.50–1.00 per 1−SD increase). Adding these phospholipids to a model including multiple predictors of dementia led to only minimal improvement in detection (C statistic changed from 0.702 to 0.71). DiscussionSome phospholipids and metabolites were altered in MCI and dementia but cross-sectional association was relatively weak and did not improve detection of MCI and dementia beyond information provided by clinical variables.