Towards Machine-FAIR: Representing software and datasets to facilitate reuse and scientific discovery by machines

OBJECTIVE

To use software, datasets, and data formats in the domain of Infectious Disease Epidemiology as a test collection to evaluate a novel M1 use case, which we introduce in this paper. M1 is a machine that upon receipt of a new digital object of research, exhaustively finds all valid compositions of it with existing objects.

METHOD

We implemented a data-format-matching-only M1 using exhaustive search, which we refer to as M1DFM. We then ran M1DFM on the test collection and used error analysis to identify needed semantic constraints.

RESULTS

Precision of M1DFM search was 61.7%. Error analysis identified needed semantic constraints and needed changes in handling of data services. Most semantic constraints were simple, but one data format was sufficiently complex to be practically impossible to represent semantic constraints over, from which we conclude limitatively that software developers will have to meet the machines halfway by engineering software whose inputs are sufficiently simple that their semantic constraints can be represented, akin to the simple APIs of services. We summarize these insights as M1-FAIR guiding principles for composability and suggest a roadmap for progressively capable devices in the service of reuse and accelerated scientific discovery.

CONCLUSION

Algorithmic search of digital repositories for valid workflow compositions has potential to accelerate scientific discovery but requires a scalable solution to the problem of knowledge acquisition about semantic constraints on software inputs. Additionally, practical limitations on the logical complexity of semantic constraints must be respected, which has implications for the design of software.

Development and validation of a race-agnostic computable phenotype for kidney health in adult hospitalized patients

Standard race adjustments for estimating glomerular filtration rate (GFR) and reference creatinine can yield a lower acute kidney injury (AKI) and chronic kidney disease (CKD) prevalence among African American patients than non-race adjusted estimates. We developed two race-agnostic computable phenotypes that assess kidney health among 139,152 subjects admitted to the University of Florida Health between 1/2012-8/2019 by removing the race modifier from the estimated GFR and estimated creatinine formula used by the race-adjusted algorithm (race-agnostic algorithm 1) and by utilizing 2021 CKD-EPI refit without race formula (race-agnostic algorithm 2) for calculations of the estimated GFR and estimated creatinine. We compared results using these algorithms to the race-adjusted algorithm in African American patients. Using clinical adjudication, we validated race-agnostic computable phenotypes developed for preadmission CKD and AKI presence on 300 cases. Race adjustment reclassified 2,113 (8%) to no CKD and 7,901 (29%) to a less severe CKD stage compared to race-agnostic algorithm 1 and reclassified 1,208 (5%) to no CKD and 4,606 (18%) to a less severe CKD stage compared to race-agnostic algorithm 2. Of 12,451 AKI encounters based on race-agnostic algorithm 1, race adjustment reclassified 591 to No AKI and 305 to a less severe AKI stage. Of 12,251 AKI encounters based on race-agnostic algorithm 2, race adjustment reclassified 382 to No AKI and 196 (1.6%) to a less severe AKI stage. The phenotyping algorithm based on refit without race formula performed well in identifying patients with CKD and AKI with a sensitivity of 100% (95% confidence interval [CI] 97%-100%) and 99% (95% CI 97%-100%) and a specificity of 88% (95% CI 82%-93%) and 98% (95% CI 93%-100%), respectively. Race-agnostic algorithms identified substantial proportions of additional patients with CKD and AKI compared to race-adjusted algorithm in African American patients. The phenotyping algorithm is promising in identifying patients with kidney disease and improving clinical decision-making.

Temporal trends in critical care utilization and outcomes in allogeneic hematopoietic stem cell transplant recipients

Historically, the prognosis of allogeneic hematopoietic stem cell transplant (allo-HCT) recipients who require intensive care unit (ICU) admission has been poor. We aimed to describe the epidemiological trends of ICU utilization and outcomes in allo-HCT patients. We conducted a retrospective cohort study including adults (≥ 18) undergoing allo-HCT between 01/01/2005 and 31/12/2020 at Mayo Clinic, Rochester. Temporal trends in outcomes were assessed by robust linear regression modelling. Risk factors for hospital mortality were chosen a priori and assessed with multivariable logistic regression modelling. Of 1,249 subjects, there were 486 ICU admissions among 287 individuals. Although older patients underwent allo-HCT (1.64 [95% CI: 1.11 to 2.45] years per year; P = 0.025), there was no change in ICU utilization over time (P = 0.91). The ICU and hospital mortality rates were 19.2% (55/287) and 28.2% (81/287), respectively. There was a decline in ICU mortality (-0.38% [95% CI: -0.70 to -0.06%] per year; P = 0.035). The 1-year post-HCT mortality for those requiring ICU admission was 56.1% (161/287), with no significant difference over time, versus 15.8% (141/891, 71 missing) among those who did not. The frequency and duration of invasive mechanical ventilation (IMV) declined. In multivariable analyses, higher serum lactate, higher sequential organ failure assessment (SOFA) scores, acute respiratory distress (ARDS), and need for IMV were associated with greater odds of hospital mortality. Over time, rates of ICU utilization have remained stable, despite increasing patient age. Several trends suggest improvement in outcomes, notably lower ICU mortality and frequency of IMV. However, long-term survival remains unchanged. Further work is needed to improve long-term outcomes in this population.

Venetoclax and hypomethylating agent combination therapy in newly diagnosed acute myeloid leukemia: Genotype signatures for response and survival among 301 consecutive patients

Venetoclax + hypomethylating agent (Ven-HMA) is currently the standard frontline therapy for older/unfit patients with newly diagnosed acute myeloid leukemia (ND-AML). Our objective in the current retrospective study of 301 adult patients (median age 73 years; 62% de novo) with ND-AML was to identify molecular predictors of treatment response to Ven-HMA and survival; European LeukemiaNet (ELN) genetic risk assignment was favorable 15%, intermediate 16%, and adverse 69%. Complete remission, with (CR) or without (CRi), count recovery, was documented in 182 (60%) patients. In multivariable analysis, inclusive of mutations only, "favorable" predictors of CR/CRi were NPM1 (86% vs. 56%), IDH2 (80% vs. 58%), and DDX41 (100% vs. 58%) and "unfavorable" TP53 (40% vs. 67%), FLT3-ITD (36% vs. 63%), and RUNX1 (44% vs. 64%) mutations; significance was sustained for each mutation after adjustment for age, karyotype, and therapy-related qualification. CR/CRi rates ranged from 36%, in the presence of unfavorable and absence of favorable mutation, to 91%, in the presence of favorable and absence of unfavorable mutation. At median follow-up of 8.5 months, 174 deaths and 41 allogeneic stem cell transplants (ASCT) were recorded. In multivariable analysis, risk factors for inferior survival included failure to achieve CR/CRi (HR 3.4, 95% CI 2.5-4.8), adverse karyotype (1.6, 1.1-2.6), TP53 mutation (1.6, 1.0-2.4), and absence of IDH2 mutation (2.2, 1.0-4.7); these risk factors were subsequently applied to construct an HR-weighted risk model that performed better than the ELN genetic risk model (AIC 1661 vs. 1750): low (n = 130; median survival 28.9 months), intermediate (n = 105; median 9.6 months), and high (n = 66; median 3.1 months; p < .001); survival in each risk category was significantly upgraded by ASCT. The current study identifies genotype signatures for predicting response and proposes a 3-tiered, CR/CRi-based, and genetics-enhanced survival model for AML patients receiving upfront therapy with Ven-HMA.

Venetoclax duration (14 vs. 21 vs. 28 days) in combination with hypomethylating agent in newly diagnosed acute myeloid leukemia: Comparative analysis of response, toxicity, and survival

Overall survival and response rates of 270 patients with newly diagnosed acute myeloid leukemia receiving venetoclax (Ven) plus hypomethylating agent, stratified by Ven dosing schedule (Cycle 1 Ven 14 vs. 21 vs. 28 days).

Molecular predictors of response and survival following IDH1/2 inhibitor monotherapy in acute myeloid leukemia

Not available.

A spatial and contextual exposome-wide association study and polyexposomic score of COVID-19 hospitalization

Environmental exposures have been linked to COVID-19 severity. Previous studies examined very few environmental factors, and often only separately without considering the totality of the environment, or the exposome. In addition, existing risk prediction models of severe COVID-19 predominantly rely on demographic and clinical factors. To address these gaps, we conducted a spatial and contextual exposome-wide association study (ExWAS) and developed polyexposomic scores (PES) of COVID-19 hospitalization leveraging rich information from individuals' spatial and contextual exposome. Individual-level electronic health records of 50 368 patients aged 18 years and older with a positive SARS-CoV-2 PCR/Antigen lab test or a COVID-19 diagnosis between March 2020 and October 2021 were obtained from the OneFlorida+ Clinical Research Network. A total of 194 spatial and contextual exposome factors from 10 data sources were spatiotemporally linked to each patient based on geocoded residential histories. We used a standard two-phase procedure in the ExWAS and developed and validated PES using gradient boosting decision trees models. Four exposome measures significantly associated with COVID-19 hospitalization were identified, including 2-chloroacetophenone, low food access, neighborhood deprivation, and reduced access to fitness centers. The initial prediction model in all patients without considering exposome factors had a testing-area under the curve (AUC) of 0.778. Incorporation of exposome data increased the testing-AUC to 0.787. Similar findings were observed in subgroup analyses focusing on populations without comorbidities and aged 18-24 years old. This spatial and contextual exposome study of COVID-19 hospitalization confirmed previously reported risk factor but also generated novel predictors that warrant more focused evaluation.

Phase 1/1b study of azacitidine and hedgehog pathway inhibitor sonidegib in patients with myeloid neoplasms

BACKGROUND

Myeloid neoplasms (myelodysplastic syndrome [MDS], myelofibrosis, and chronic myelomonocytic [CMML]) are aggressive hematological malignancies for which, despite recent approvals, novel therapies are needed to improve clinical outcomes. The hedgehog (HH) pathway is one of the main pathways for cancer stem cells survival and several HH inhibitors (HHi) are approved in clinical practice.

METHODS

Sonidegib (SON), an oral HHi, was tested in this phase 1/1b trial in combination with azacitidine (AZA, 75 mg/m² days ×7) in patients with newly diagnosed and relapsed/refractory (r/r) chronic MN or acute myeloid leukemia (AML).

RESULTS

Sixty-two patients (28 [45%] newly diagnosed) were treated in this study, including 10 patients in the dose-finding component and 52 patients in phase 1b. SON 200 mg oral daily on days 1-28 each cycle was deemed the recommended dose for phase 1b. Out of 21 rrAML patients, two achieved response (one complete response/one morphologic leukemia-free state) with no responses seen in seven r/r MDS/CMML patients. In newly diagnosed AML/MDS, response was seen in six (three had complete remission, two had morphological leukemia-free status) of 27 patients. Median overall survival was 26.4 and 4.7 months for newly diagnosed MDS and AML, respectively. Safety was satisfactory with common (>20%) side effects including fatigue, constipation, nausea, cough, insomnia, and diarrhea. Only 7% of patients died in the study, and none of the deaths were deemed related to treatment.

CONCLUSIONS

Our study shows that AZA + SON are a safe combination in a patient with MN. Similar to other hedgehog inhibitors, this combination yielded limited response rate in patients with myeloid neoplasms.

Abstract P394: Initial Antihypertensive Prescribing in Relation to Blood Pressure Among Medicaid and Medicare Recipients

Introduction: Early treatment for hypertension (HTN) portends better outcomes. However, few real-world studies have assessed initial antiHTN regimens and how they differ by baseline blood pressure (BP). We sought to compare initial treatment patterns, stratified by BP, between Medicaid and Medicare recipients.

Methods: We performed a cross-sectional study of adults with newly-treated HTN in the One Florida+ Consortium(2012-2020) who had linked claims-EHR data from the treatment initiation visit. Eligible patients were Floridians with Medicaid or Medicare aged ≥18 years, with diagnosed HTN, who filled ≥1 first-line antiHTN class with no evidence of anti HTN fills during the year prior (in which continuous insurance enrollment was required). Baseline BP was categorized per current HTN guidelines, and logistic regression was used to estimate age-adjusted odds of combination vs. monotherapy, per 10 mmHg increase in systolic BP (SBP) or diastolic BP (DBP).

Results: We included 2,902 patients (47% Medicaid, 53% Medicare); mean age was 44 (Medicaid) and 67 yrs(Medicare); 60% (64% and 56%, respectively) were women and 42% (57% and 29%, respectively) were Black. Initial antiHTN classes were similar comparing cohorts: ACEI, ARB and thiazide initiation varied little by BP category, in contrast to CCBs and β-blockers (Figure, panels A-B). In age-adjusted analyses, use of initial combination therapy was 40% more likely (OR, 1.40; 95% CI, 1.11, 1.76) among Medicare recipients and inversely related to BP category (panels C-D) among Medicare patients, in which each 10mmHg greater SBP (OR, 0.93; 95% CI, 0.88, 0.97) and DBP (OR 0.82; 95% CI, 0.75, 0.90) had lower odds of combination therapy. Among Medicaid recipients, only SBP associated with combination therapy (OR1.11; 95% CI, 1.03, 1.20).

Conclusions: We observed similar initial class patterns among Medicaid & Medicare recipients across baseline BP, but differential use of combination therapy was less likely at higher baseline BP in Medicare recipients, which contrasts current guidance.

Predictors of anemia response to momelotinib therapy in myelofibrosis and impact on survival

We retrospectively reviewed 72 anemic patients with myelofibrosis (MF; median age 68 years), who were JAK2 inhibitor-naïve at the time of study entry to a phase-1/2 momelotinib clinical trial. Driver mutation profile included JAK2 69%, CALR 17%, MPL 8%, and triple-negative 6%; other mutations included ASXL1 39% and SRSF2 17%. Momelotinib was administered at a median dose of 300 mg daily. Anemia response was assessed by formal criteria and documented in 44% of all patients with hemoglobin levels below the sex-adjusted reference range (n = 72), 48% of those with hemoglobin <10 g/dl (n = 54), and 46% of those who were transfusion-dependent at the time of study entry (n = 28). Anemia response was more likely with post-essential thrombocythemia MF (83% vs 37%; p = .001), lower serum ferritin (p = .003), and shorter time from diagnosis to momelotinib therapy (p = .001); the first two variables were also predictive in transfusion-dependent patients. Post-momelotinib median survival was 3.2 years; in univariate analysis, survival was superior in anemia responders (median 3.8 vs. 2.8 years; p = .14) and in the presence of type 1/like CALR mutation and inferior in the presence of age > 65 years, ASXL1/SRSF2 mutation, unfavorable karyotype, DIPSS-plus high risk, red cell transfusion need and higher serum ferritin. Multivariable analysis confirmed the favorable impact of anemia response on survival (p = .02; HR 0.5, 3/5/10-year survival; 69%/38%/25%). This survival advantage was also noted in transfusion-dependent patients (3.7 vs. 1.9 years; p = .01; HR 0.3) and appeared to be restricted to patients with an unfavorable genetic profile. The current study suggests a short-term survival benefit associated with anemia response in momelotinib-treated patients with MF.

Lack of a caregiver is associated with shorter survival in myeloma patients undergoing autologous stem cell transplantation

There is increasing evidence that social infrastructure and a healthy social network can improve cancer survival. Mayo Clinic has an outpatient stem cell transplantation program for myeloma. Safe outpatient transplantation requires a caregiver to be present. Patients lacking a caregiver are transplanted as an inpatient. We reviewed outcomes on over 2000 patients with multiple myeloma, 2103 transplanted as an outpatient compared with 41 hospitalized for transplantation. Although progression-free survival following transplantation was identical between the two groups, overall survival was shorter in those hospitalized. This suggests that the absence of a caregiver for transplantation is an important surrogate of the social infrastructure associated with poor outcomes in transplanted patients with multiple myeloma.

Abnormal CD13/HLA-DR Expression Pattern on Myeloblasts Predicts Development of Myeloid Neoplasia in Patients With Clonal Cytopenia of Undetermined Significance

OBJECTIVES

Patients with clonal cytopenia of undetermined significance (CCUS) are at increased risk of developing myeloid neoplasia (MN). We evaluated whether a simple flow cytometry immunophenotyping (FCIP) assay could differentiate the risk of development of MN in patients with CCUS.

METHODS

Bone marrow aspirates were assessed by FCIP panel in a cohort of 80 patients identified as having CCUS based on next-generation sequencing or cytogenetics from March 2015 to May 2020, with available samples. Flow cytometric assay included CD13/HLA-DR expression pattern on CD34-positive myeloblasts; CD13/CD16 pattern on maturing granulocytic precursors; and aberrant expression of CD2, CD7, or CD56 on CD34-positive myeloblasts. Relevant demographic, comorbidity, and clinical and laboratory data, including the type and extent of genetic abnormalities, were extracted from the electronic health record.

RESULTS

In total, 17 (21%) patients with CCUS developed MN over the follow-up period (median survival follow-up, 28 months [95% confidence interval, 19-31]). Flow cytometry immunophenotyping abnormalities, including the aberrant pattern of CD13/HLA-DR expression, as detected at the time of the diagnosis of CCUS, were significantly associated with risk of developing MN (hazard ratio, 2.97; P = .006). Additional FCIP parameters associated with the development of MN included abnormal expression of CD7 on myeloblasts and the presence vs absence of any FCIP abnormality.

CONCLUSIONS

A simple FCIP approach that includes assessment of CD13/HLA-DR pattern on CD34-positive myeloblasts can be useful in identifying patients with CCUS at higher risk of developing MN.

Impact of Center Experience with Donor Type on Outcomes: A Secondary Analysis, Blood and Marrow Transplant Clinical Trials Network 1101Open for Accrual June 2012Open for Accrual June 2012

We previously reported the results of Blood and Marrow Transplant Clinical Trials Network (BMT CTN) 1101, a randomized comparison of hematopoietic cell transplantation (HCT) performed with double umbilical cord blood units (dUCB) or with haploidentical bone marrow (haplo-BMT) with post-transplantation cyclophosphamide (PTCy) in the nonmyeloablative setting. Those results showed similar progression-free survival in the 2 treatment groups but lower nonrelapse mortality and better overall survival in the haplo-BM arm. In this secondary analysis, we sought to investigate whether transplantation center's previous experience with haplo-BM and/or dUCB HCT had an impact on outcomes. All patients randomized in BMT CTN 1101 were included. Center experience was assigned based on the number of transplantations with each platform performed in the year before initiation of the study according to the Center for International Blood and Marrow Transplant Research. Centers were then classified as a dUCB center (>10 dUCB HCTs; n = 117 patients, 10 centers), a haplo-BM center (>10 haplo-BM HCTs and ≤10 dUCB HCTs; n = 110 patients, 2 centers), or other center (≤10 haplo and ≤10 dUCB HCTs; n = 140 patients, 21 centers). After adjusting for age, Karnofsky Performance Status, and Disease Risk Index, we found that haplo-BM centers had lower overall mortality with this donor type compared with dUCB centers (hazard ratio [HR], 2.56; 95% confidence interval [CI], 1.44 to 4.56). In contrast, there were no differences in overall mortality between haplo-BM and dUCB in centers that were experienced with dUCB HCT (HR, 1.02; 95% CI, .59 to 1.79) or had limited to no experience with either dUCB or haplo-BM HCT (HR, 1.36; 95% CI, .83 to 2.21). The higher risk of treatment failure and overall mortality in dUCB HCT in haplo BM-experienced centers was driven by a significantly higher risk of relapse (HR, 1.78; 95% CI, 1.07 to 2.97). With the exception of worse outcomes among dUCB HCT recipients in haplo-BM centers, transplantation center experience in the year before initiation of BMT CTN 1101 had a limited impact on the outcomes of this randomized clinical trial.

Integration of Biobanking Architecture with Genomics Data: Genomics Integrated Biobanking Ontology (GIBO)

Integration of clinical-pathological information of Biobanks with genomics-epidemiological data/inferences in a structured and consistent manner, mitigating inherent heterogeneities of sites/sources of data/sample collection, processing, and information storage hurdles, is primary to achieving an automated surveillance system. Genomics Integrated Biobanking Ontology (GIBO) presents a solution for preserving the contextual meaning of heterogeneous data, while interlinking different genomics and epidemiological concepts in machine comprehensible format with the biobank framework. GIBO an OWL ontology introduces 84 new classes to integrate genomics data relevant to public health.

Impact of maintenance therapy post autologous stem cell transplantation for multiple myeloma in early and delayed transplant

Based on phase 3 trials, maintenance therapy after autologous stem cell transplantation (ASCT) has become the standard of care in multiple myeloma (MM). We examined the trends in maintenance therapy in a large group of patients (2530) transplanted at a single institution over two decades. Majority (n = 1958; 77%) had an ASCT within 12 months of diagnosis (early ASCT). Maintenance was employed in 39% of the patients; 42% among early ASCT and 30.5% among delayed ASCT. Most common maintenance approach was an IMiD (61%), followed by a PI (31%), or a PI + IMiD (4%). Patients with high-risk FISH received PI-based maintenance more frequently. The PFS was superior with maintenance (36 vs. 22 months, p < 0.001); 37 vs. 25 months for early ASCT (p < 0.001) and 29 vs. 17 months for delayed ASCT (p = 0.0008). OS from ASCT was higher with maintenance for the whole cohort at 93 vs. 73 months (p < 0.001). OS from diagnosis was also better for the whole cohort with maintenance therapy, 112 vs. 93 months (p < 0.001). The improvement in PFS and OS was seen in high-risk and standard risk disease. The experience with maintenance therapy post ASCT for myeloma in a non-clinical trial setting confirms the findings from the phase 3 trials.

Extracting social determinants of health from electronic health records using natural language processing: a systematic review

OBJECTIVE

Social determinants of health (SDoH) are nonclinical dispositions that impact patient health risks and clinical outcomes. Leveraging SDoH in clinical decision-making can potentially improve diagnosis, treatment planning, and patient outcomes. Despite increased interest in capturing SDoH in electronic health records (EHRs), such information is typically locked in unstructured clinical notes. Natural language processing (NLP) is the key technology to extract SDoH information from clinical text and expand its utility in patient care and research. This article presents a systematic review of the state-of-the-art NLP approaches and tools that focus on identifying and extracting SDoH data from unstructured clinical text in EHRs.

MATERIALS AND METHODS

A broad literature search was conducted in February 2021 using 3 scholarly databases (ACL Anthology, PubMed, and Scopus) following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. A total of 6402 publications were initially identified, and after applying the study inclusion criteria, 82 publications were selected for the final review.

RESULTS

Smoking status (n = 27), substance use (n = 21), homelessness (n = 20), and alcohol use (n = 15) are the most frequently studied SDoH categories. Homelessness (n = 7) and other less-studied SDoH (eg, education, financial problems, social isolation and support, family problems) are mostly identified using rule-based approaches. In contrast, machine learning approaches are popular for identifying smoking status (n = 13), substance use (n = 9), and alcohol use (n = 9).

CONCLUSION

NLP offers significant potential to extract SDoH data from narrative clinical notes, which in turn can aid in the development of screening tools, risk prediction models, and clinical decision support systems.

Evaluation of Elafin as a Prognostic Biomarker in Acute Graft-versus-Host Disease

Acute graft-versus-host disease (GVHD) is a major cause of mortality in patients undergoing hematopoietic cell transplantation (HCT) for hematologic malignancies. The skin is the most commonly involved organ in GVHD. Elafin, a protease inhibitor overexpressed in inflamed epidermis, was previously identified as a diagnostic biomarker of skin GVHD; however, this finding was restricted to a subset of patients with isolated skin GVHD. The main driver of nonrelapse mortality (NRM) in HCT recipients is gastrointestinal (GI) GVHD. Two biomarkers, Regenerating islet-derived 3a (REG3α) and Suppressor of tumorigenesis 2 (ST2), have been validated as biomarkers of GI GVHD that predict long-term outcomes in patients treated for GVHD. We undertook this study to determine the utility of elafin as a prognostic biomarker in the general population of acute GVHD patients in whom GVHD may develop in multiple organs. We analyzed serum elafin concentrations as a predictive biomarker of acute GVHD outcomes and compared it with ST2 and REG3α in a large group of patients treated at multiple centers. A total of 526 patients from the Mount Sinai Acute GVHD International Consortium (MAGIC) who had received corticosteroid treatment for skin GVHD and who had not been previously studied were analyzed. Serum concentrations of elafin, ST2, and REG3α were measured by ELISA in all patients. The patients were divided at random into equal training and validation sets, and a competing-risk regression model was developed to model 6-month NRM using elafin concentration in the training set. Additional models were developed using concentrations of ST2 and REG3α or the combination of all 3 biomarkers as predictors. Receiver operating characteristic (ROC) curves were constructed using the validation set to evaluate the predictive accuracy of each model and to stratify patients into high- and low-risk biomarker groups. The cumulative incidence of 6-month NRM, overall survival (OS), and 4-week treatment response were compared between the risk groups. Unexpectedly, patients in the low-risk elafin group demonstrated a higher incidence of 6-month NRM, although the difference was not statistically significant (17% versus 11%; P = .19). OS at 6 months (68% versus 68%; P > .99) and 4-week response (78% versus 78%; P = .98) were similar in the low-risk and high-risk elafin groups. The area under the ROC curve (AUC) was 0.55 for elafin and 0.75 for the combination of ST2 and REG3α. The addition of elafin to the other 2 biomarkers did not improve the AUC. Our data indicate that serum elafin concentrations measured at the initiation of systemic treatment for acute GVHD did not predict 6-month NRM, OS, or treatment response in a multicenter population of patients treated systemically for acute GVHD. As seen in previous studies, serum concentrations of the GI GVHD biomarkers ST2 and REG3α were significant predictors of NRM, and the addition of elafin levels did not improve their accuracy. These results underscore the importance of GI disease in driving NRM in patients who develop acute GVHD.

Ruxolitinib resistance or intolerance in steroid-refractory acute graft-versus-host disease - a real-world outcomes analysis

Ruxolitinib for steroid-refractory acute graft-versus-host disease (SR-aGVHD) results in resistance or intolerance in 1/5 of patients. Outcomes of such patients are undefined. We identified these patients in a multicentre review and reported outcomes. Ruxolitinib-resistant aGVHD was identified in 48/307 patients. Among patients receiving additional therapy, the overall response rate to next therapy was 36%. Median survival was 21 days. Ruxolitinib intolerance led to treatment discontinuation in 16/307 patients. Ten intolerant patients received additional therapy with 50% experiencing continued improvement of aGVHD. Median survival was 50 days in these patients. These data serve as a baseline for future SR-aGVHD studies.

Assessing the practice of data quality evaluation in a national clinical data research network through a systematic scoping review in the era of real-world data

Objective

To synthesize data quality (DQ) dimensions and assessment methods of real-world data, especially electronic health records, through a systematic scoping review and to assess the practice of DQ assessment in the national Patient-centered Clinical Research Network (PCORnet).

Materials and Methods

We started with 3 widely cited DQ literature—2 reviews from Chan et al (2010) and Weiskopf et al (2013a) and 1 DQ framework from Kahn et al (2016)—and expanded our review systematically to cover relevant articles published up to February 2020. We extracted DQ dimensions and assessment methods from these studies, mapped their relationships, and organized a synthesized summarization of existing DQ dimensions and assessment methods. We reviewed the data checks employed by the PCORnet and mapped them to the synthesized DQ dimensions and methods.

Results

We analyzed a total of 3 reviews, 20 DQ frameworks, and 226 DQ studies and extracted 14 DQ dimensions and 10 assessment methods. We found that completeness, concordance, and correctness/accuracy were commonly assessed. Element presence, validity check, and conformance were commonly used DQ assessment methods and were the main focuses of the PCORnet data checks.

Discussion

Definitions of DQ dimensions and methods were not consistent in the literature, and the DQ assessment practice was not evenly distributed (eg, usability and ease-of-use were rarely discussed). Challenges in DQ assessments, given the complex and heterogeneous nature of real-world data, exist.

Conclusion

The practice of DQ assessment is still limited in scope. Future work is warranted to generate understandable, executable, and reusable DQ measures.

Developing and Validating a Computable Phenotype for the Identification of Transgender and Gender Nonconforming Individuals and Subgroups

Transgender and gender nonconforming (TGNC) individuals face significant marginalization, stigma, and discrimination. Under-reporting of TGNC individuals is common since they are often unwilling to self-identify. Meanwhile, the rapid adoption of electronic health record (EHR) systems has made large-scale, longitudinal real-world clinical data available to research and provided a unique opportunity to identify TGNC individuals using their EHRs, contributing to a promising routine health surveillance approach. Built upon existing work, we developed and validated a computable phenotype (CP) algorithm for identifying TGNC individuals and their natal sex (i.e., male-to-female or female-to-male) using both structured EHR data and unstructured clinical notes. Our CP algorithm achieved a 0.955 F1-score on the training data and a perfect F1-score on the independent testing data. Consistent with the literature, we observed an increasing percentage of TGNC individuals and a disproportionate burden of adverse health outcomes, especially sexually transmitted infections and mental health distress, in this population.

International Classification of Diseases, Tenth Revision, Clinical Modification social determinants of health codes are poorly used in electronic health records

There have been increasing calls for clinicians to document social determinants of health (SDOH) in electronic health records (EHRs). One potential source of SDOH in the EHRs is in the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) Z codes (Z55-Z65). In February 2018, ICD-10-CM Official Guidelines for Coding and Reporting approved that all clinicians, not just the physicians, involved in the care of a patient can document SDOH using these Z codes.To examine the utilization rate of the ICD-10-CM Z codes using data from a large network of EHRs.We conducted a retrospective analysis of EHR data between 2015 to 2018 in the OneFlorida Clinical Research Consortium, 1 of the 13 Clinical Data Research Networks funded by Patient-Centered Outcomes Research Institute. We calculated the Z code utilization rate at both the encounter and patient levels.We found a low rate of utilization for these Z codes (270.61 per 100,000 at the encounter level and 2.03% at the patient level). We also found that the rate of utilization for these Z codes increased (from 255.62 to 292.79 per 100,000) since the official approval of Z code reporting from all clinicians by the American Hospital Association Coding Clinic and ICD-10-CM Official Guidelines for Coding and Reporting became effective in February 2018.The SDOH Z codes are rarely used by clinicians. Providing clear guidelines and incentives for documenting the Z codes can promote their use in EHRs. Improvements in the EHR systems are probably needed to better document SDOH.

Applications of artificial intelligence in drug development using real-world data

The US Food and Drug Administration (FDA) has been actively promoting the use of real-world data (RWD) in drug development. RWD can generate important real-world evidence reflecting the real-world clinical environment where the treatments are used. Meanwhile, artificial intelligence (AI), especially machine- and deep-learning (ML/DL) methods, have been increasingly used across many stages of the drug development process. Advancements in AI have also provided new strategies to analyze large, multidimensional RWD. Thus, we conducted a rapid review of articles from the past 20 years, to provide an overview of the drug development studies that use both AI and RWD. We found that the most popular applications were adverse event detection, trial recruitment, and drug repurposing. Here, we also discuss current research gaps and future opportunities.

Autologous stem cell transplantation for multiple myeloma patients aged ≥ 75 treated with novel agents

Autologous stem cell transplantation (ASCT) has been used for treating multiple myeloma (MM) for over three decades and is generally reserved for patients younger than 65. Herein we report on outcomes of outpatient ASCT in a cohort of patients with MM aged ≥75 years. Between October 2005 and August 2020, 50 patients aged ≥75 years, received an ASCT at Mayo Clinic, Rochester. Median time from diagnosis to ASCT was 6.85 months (IQR 5.2-10.52) and 50%. received reduced intensity conditioning with melphalan 140 mg/m². 48% of patients completed the ASCT without requiring hospitalization and 52% (n = 26) of patients required hospitalization with a median duration of hospital admission of 9 days (IQR 5-13). Reasons for hospitalization included fever or infection (32%), cardiac arrhythmia (36%), and dehydration (32%). Overall response rate was 100% with a complete response seen in 57% of patients. Median overall survival and progression free survival for the cohort were 82 months and 33 months, respectively. One patient died within 100 days of transplant representing a 2% 100-day mortality rate. ASCT is safe and efficacious in carefully selected MM patients aged 75 or above and we believe that age should not be an exclusion factor for ASCT in MM.

Venetoclax and hypomethylating agents in acute myeloid leukemia: Mayo Clinic series on 86 patients

Venetoclax and hypomethylating agent (HMA) combination therapy is FDA-approved for elderly or unfit acute myeloid leukemia (AML) patients unable to withstand intensive chemotherapy. The primary objective of the current study was to impart our institutional experience with the above regimen, outlining response, survival outcomes, and its determinants amongst 86 treatment- naïve and relapsed/refractory AML patients. A total of 44 treatment-naïve AML patients, median age 73.5 years, enriched with secondary, therapy related and ELN adverse risk disease (n = 27) were studied. The CR/CRi rates of 50% (22 of 44 patients) were superior to 23% in a matched AML cohort treated with HMA alone (P = .005). Response rates were similar with TP53, FLT3, NPM1 and IDH mutations (P = .31). Moreover, CEPBA mutations (P = .03) and neutropenia (P = .05) emerged as predictors of complete response. Survivalwas prolonged in patients achieving CR/CRi (17 vs 3 months without CR/CRi, P < .001; conversely adverse ELN risk portended inferior survival. Amongst 42 relapsed/refractory AML patients, half received ≥2 prior therapies excluding transplant, and 15 (35.7%) had received HMA. A group of 14 patients (33.3%) attained CR/CRi; age > 65 years, AML with myelodysplasia, JAK2, DNMT3A, and BCOR mutations predicted complete response. Survival distinctions were based on CR/CRi (median survival 15 vs 3 months with/without CR/CRi; P < .001), and TP53 mutation status (P = .04). In summary, we corroborate existing reports demonstrating superior response and prolonged survival with venetoclax and HMA in treatment -naïve and relapsed/refractory AML patients regardless of genotype. Additionally, we identify unique predictors of response to therapy which require validation.

Depth of response prior to autologous stem cell transplantation predicts survival in light chain amyloidosis

The goal of therapy in AL amyloidosis is to inhibit further production of the amyloidogenic light chains, thereby allowing organ recovery and improving survival. We aimed to assess the impact of depth of hematologic response prior to ASCT on survival. We conducted a retrospective study of 128 newly diagnosed AL amyloidosis patients who received induction prior to ASCT between January 2007 and August 2017 at Mayo Clinic. The overall response rate to induction was 86% (CR 18%, VGPR 31% and PR 38%). With a median follow up of 52 months, the median PFS and OS was 48.5 months and not reached, respectively. Response depth to induction therapy was associated with improved PFS and OS. The median PFS was not reached for patients achieving ≥VGPR prior to ASCT and 34.1 months for patients achieving PR or less (P = 0.0009). The median OS was longer in patients with deeper responses (not reached for ≥VGPR vs. 128 months for PR or less (P = 0.02)). On multivariable analysis, independent predictors of OS were melphalan conditioning dose (RR = 0.42; P = 0.036) and depth of response prior to transplant (RR 0.37; P = 0.0295). Hematologic response prior to transplant predicts improved post transplant outcomes in AL amyloidosis.

Outcomes with early vs. deferred stem cell transplantation in light chain amyloidosis

In the presence of effective treatment options for systemic light chain (AL) amyloidosis, autologous stem cell transplantation (ASCT) is sometimes deferred after stem cell collection. We designed this retrospective study to compare overall survival (OS) between patients who proceed directly to ASCT after stem cell collection and those who defer ASCT. We included patients with AL amyloidosis who had stem cell collection at Mayo Clinic, Minnesota, from 2004 to 2018. ASCT was considered "early" if performed within 90 days of collection, and "deferred" if performed after 90 days, or not done by last follow up. We included 651 patients; 527 underwent early ASCT and 124 deferred ASCT. There was no difference in OS with early vs. deferred ASCT (median OS: 13.0 vs. 11.4 years, respectively, P = 0.28). There was no difference in OS between the 2 groups among patients with early or advanced Mayo Stage. Among patients who achieved ≥very good partial response at the time of collection, OS in the early and deferred groups was 14.2 and 13.4 years, respectively (P = 0.06). Survival outcomes are similar with early and deferred ASCT. Further studies are needed to identify patients who would benefit from each approach.

Clinical Trial Generalizability Assessment in the Big Data Era: A Review

Clinical studies, especially randomized, controlled trials, are essential for generating evidence for clinical practice. However, generalizability is a long‐standing concern when applying trial results to real‐world patients. Generalizability assessment is thus important, nevertheless, not consistently practiced. We performed a systematic review to understand the practice of generalizability assessment. We identified 187 relevant articles and systematically organized these studies in a taxonomy with three dimensions: (i) data availability (i.e., before or after trial (a priori vs. a posteriori generalizability)); (ii) result outputs (i.e., score vs. nonscore); and (iii) populations of interest. We further reported disease areas, underrepresented subgroups, and types of data used to profile target populations. We observed an increasing trend of generalizability assessments, but < 30% of studies reported positive generalizability results. As a priori generalizability can be assessed using only study design information (primarily eligibility criteria), it gives investigators a golden opportunity to adjust the study design before the trial starts. Nevertheless, < 40% of the studies in our review assessed a priori generalizability. With the wide adoption of electronic health records systems, rich real‐world patient databases are increasingly available for generalizability assessment; however, informatics tools are lacking to support the adoption of generalizability assessment practice.

Assessing the Validity of a a priori Patient-Trial Generalizability Score using Real-world Data from a Large Clinical Data Research Network: A Colorectal Cancer Clinical Trial Case Study

Existing trials had not taken enough consideration of their population representativeness, which can lower the effectiveness when the treatment is applied in real-world clinical practice. We analyzed the eligibility criteria of Bevacizumab colorectal cancer treatment trials, assessed their a priori generalizability, and examined how it affects patient outcomes when applied in real-world clinical settings. To do so, we extracted patient-level data from a large collection of electronic health records (EHRs) from the OneFlorida consortium. We built a zero-inflated negative binomial model using a composite patient-trial generalizability (cPTG) score to predict patients' clinical outcomes (i.e., number of serious adverse events, [SAEs]). Our study results provide a body of evidence that 1) the cPTG scores can predict patient outcomes; and 2) patients who are more similar to the study population in the trials that were used to develop the treatment will have a significantly lower possibility to experience serious adverse events.

Alternative donor transplantation for acute myeloid leukemia in patients aged ≥50 years: young HLA-matched unrelated or haploidentical donor?

We sought to study whether survival after haploidentical transplantation is comparable to that after matched unrelated donor transplantation for 822 patients aged 50-75 years with acute myeloid leukemia in first or second complete remission. One hundred and ninety-two patients received grafts from haploidentical donors (sibling 25%; offspring 75%) and 631 patients from matched unrelated donors aged 18-40 years. Patients’ and disease characteristics of the two groups were similar except that recipients of matched unrelated donor transplantation were more likely to have poor risk cytogenetics and more likely to receive myeloablative conditioning regimens. Time from documented remission to transplant did not differ by donor type. Five-year overall survival was 32% and 42% after haploidentical and matched unrelated donor transplant, respectively (P=0.04). Multivariable analysis showed higher mortality (hazard ratio 1.27, P=0.04) and relapse (hazard ratio 1.32, P=0.04) after haploidentical transplantation, with similar non-relapse mortality risks. Chronic graft-versus-host disease was higher after matched unrelated donor compared to haploidentical transplantation when bone marrow was the graft (hazard ratio 3.12, P<0.001), but when the graft was peripheral blood, there was no difference in the risk of chronic graft-versus-host disease between donor types. These data support the view that matched unrelated donor transplant with donors younger than 40 years is to be preferred.

Implementing a hash-based privacy-preserving record linkage tool in the OneFlorida clinical research network

Objective

To implement an open-source tool that performs deterministic privacy-preserving record linkage (RL) in a real-world setting within a large research network.

Materials and Methods

We learned 2 efficient deterministic linkage rules using publicly available voter registration data. We then validated the 2 rules’ performance with 2 manually curated gold-standard datasets linking electronic health records and claims data from 2 sources. We developed an open-source Python-based tool—OneFL Deduper—that (1) creates seeded hash codes of combinations of patients’ quasi-identifiers using a cryptographic one-way hash function to achieve privacy protection and (2) links and deduplicates patient records using a central broker through matching of hash codes with a high precision and reasonable recall.

Results

We deployed the OneFl Deduper (https://github.com/ufbmi/onefl-deduper) in the OneFlorida, a state-based clinical research network as part of the national Patient-Centered Clinical Research Network (PCORnet). Using the gold-standard datasets, we achieved a precision of 97.25∼99.7% and a recall of 75.5%. With the tool, we deduplicated ∼3.5 million (out of ∼15 million) records down to 1.7 million unique patients across 6 health care partners and the Florida Medicaid program. We demonstrated the benefits of RL through examining different disease profiles of the linked cohorts.

Conclusions

Many factors including privacy risk considerations, policies and regulations, data availability and quality, and computing resources, can impact how a RL solution is constructed in a real-world setting. Nevertheless, RL is a significant task in improving the data quality in a network so that we can draw reliable scientific discoveries from these massive data resources.

Etiologies of Extreme Thrombocytosis: A Contemporary Series

OBJECTIVE

To describe the multifactorial etiologies of extreme thrombocytosis (EXT) in different care settings and the frequency of finding an occult malignancy.

PATIENTS AND METHODS

We conducted a retrospective chart review at Mayo Clinic from January 1, 2011, through December 31, 2016. Adult patients who had at least 2 readings of platelet counts greater than 1000×10⁹/L within 30 days of each other were included. We determined the causes of EXT on the basis of preset definitions of precipitating factors and identified the dominant causes on the basis of the trend of platelet counts.

RESULTS

A total of 44,490 patients had thrombocytosis, and 305 patients (0.7%) had EXT. In 242 patients (79.3%), EXT was multifactorial. Surgical complications (54.1%) and hematologic malignancies (27.9%) were the 2 most dominant causes. Thirty-eight patients (12.5%) had new diagnoses of malignancies, mostly myeloproliferative neoplasms. In inpatients, surgical complications (71.9%), concurrent/previous splenectomy (50.5%), and infections (44.9%) were the most common causes, whereas hematologic malignancies (56.9%), iron deficiency (36.7%), and previous splenectomy (28.4%) were the most common causes in outpatients. Hematologic malignancy was 3.4 times more likely to be the cause of EXT in outpatients than in inpatients (56.9% vs 16.8%), and a new diagnosis of hematologic malignancy was 1.9 times more likely to be made in outpatients (15.6% vs 8.2%). Eighty-four percent of patients had resolution of EXT within 30 days. One patient died during the period of EXT. Nonsurgical patients with hematologic malignancies had the most prolonged period of EXT.

CONCLUSION

Extreme thrombocytosis is a multifactorial hematologic condition, and its etiology differs substantially between inpatients and outpatients. Occult hematologic malignancies are uncommon in EXT when other major causes are present.

Prognostic Score and Cytogenetic Risk Classification for Chronic Lymphocytic Leukemia Patients: Center for International Blood and Marrow Transplant Research Report

PURPOSE

To develop a prognostic model and cytogenetic risk classification for previously treated patients with chronic lymphocytic leukemia (CLL) undergoing reduced intensity conditioning (RIC) allogeneic hematopoietic cell transplantation (HCT).

EXPERIMENTAL DESIGN

We performed a retrospective analysis of outcomes of 606 patients with CLL who underwent RIC allogeneic HCT between 2008 and 2014 reported to the Center for International Blood and Marrow Transplant Research.

RESULTS

On the basis of multivariable models, disease status, comorbidity index, lymphocyte count, and white blood cell count at HCT were selected for the development of prognostic model. Using the prognostic score, we stratified patients into low-, intermediate-, high-, and very-high-risk [4-year progression-free survival (PFS) 58%, 42%, 33%, and 25%, respectively, P < 0.0001; 4-year overall survival (OS) 70%, 57%, 54%, and 38%, respectively, P < 0.0001]. We also evaluated karyotypic abnormalities together with del(17p) and found that del(17p) or ≥5 abnormalities showed inferior PFS. Using a multivariable model, we classified cytogenetic risk into low, intermediate, and high (P < 0.0001). When the prognostic score and cytogenetic risk were combined, patients with low prognostic score and low cytogenetic risk had prolonged PFS (61% at 4 years) and OS (75% at 4 years).

CONCLUSIONS

In this large cohort of patients with previously treated CLL who underwent RIC HCT, we developed a robust prognostic scoring system of HCT outcomes and a novel cytogenetic-based risk stratification system. These prognostic models can be used for counseling patients, comparing data across studies, and providing a benchmark for future interventions. For future study, we will further validate these models for patients receiving targeted therapies prior to HCT.

Clinical outcome of patients diagnosed with myelodysplastic syndrome-unclassifiable (MDS-U): single center experience

Myelodysplastic syndrome unclassifiable (MDS-U) is a small subtype of myelodysplastic syndromes (MDS). However, rare literature exists in terms of natural progression and clinical outcome of patients with MDS-U. In the present study, we investigated the characteristics and the clinical outcomes of patients categorized as MDS-U based on 2008 World Health Organization criteria (WHO) in a single center comparing to other MDS groups. Out of eight hundred and two patients who met WHO criteria for MDS at our institution, ninety patients (11%) were initially classified as MDS-U. Upon pathological review, only half of the cases were confirmed to be MDS-U. With follow up, half of the MDS-U cases were reclassified to another subtype. We found neither significant difference in median overall survival nor in risk of transformation to acute myeloid leukemia when comparing MDS-U to other MDS groups. Additional larger studies are needed to confirm our results.