1. Medicine
Download icon

Inference from longitudinal laboratory tests characterizes temporal evolution of COVID-19-associated coagulopathy (CAC)

  1. Colin Pawlowski
  2. Tyler Wagner
  3. Arjun Puranik
  4. Karthik Murugadoss
  5. Liam Loscalzo
  6. AJ Venkatakrishnan
  7. Rajiv K Pruthi
  8. Damon E Houghton
  9. John C O'Horo
  10. William G Morice II
  11. Amy W Williams
  12. Gregory J Gores
  13. John Halamka
  14. Andrew D Badley
  15. Elliot S Barnathan
  16. Hideo Makimura
  17. Najat Khan
  18. Venky Soundararajan  Is a corresponding author
  1. nference, inc, United States
  2. Mayo Clinic, United States
  3. Mayo Clinic Laboratories, United States
  4. Mayo Clinic Platform, United States
  5. Janssen pharmaceutical companies of Johnson & Johnson (J&J), United States
Research Article
  • Cited 0
  • Views 480
  • Annotations
Cite this article as: eLife 2020;9:e59209 doi: 10.7554/eLife.59209

Abstract

Temporal inference from laboratory testing results and triangulation with clinical outcomes extracted from unstructured electronic health record (EHR) provider notes is integral to advancing precision medicine. Here, we studied 246 SARS-CoV-2 PCR-positive (COVIDpos) patients and propensity-matched 2460 SARS-CoV-2 PCR-negative (COVIDneg) patients subjected to around 700,000 lab tests cumulatively across 194 assays. Compared to COVIDneg patients at the time of diagnostic testing, COVIDpos patients tended to have higher plasma fibrinogen levels and lower platelet counts. However, as the infection evolves, COVIDpos patients distinctively show declining fibrinogen, increasing platelet counts, and lower white blood cell counts. Augmented curation of EHRs suggests that only a minority of COVIDpos patients develop thromboembolism, and rarely, disseminated intravascular coagulopathy (DIC), with patients generally not displaying platelet reductions typical of consumptive coagulopathies. These temporal trends provide fine-grained resolution into COVID-19 associated coagulopathy (CAC) and set the stage for personalizing thromboprophylaxis.

Introduction

There is a growing body of evidence suggesting that severe COVID-19 outcomes may be associated with dysregulated coagulation (Tang et al., 2020), including stroke, pulmonary embolism, myocardial infarction, and other venous or arterial thromboembolic complications (Klok et al., 2020). This so-called COVID-19 associated coagulopathy (CAC) shares similarities with disseminated intravascular coagulation (DIC) and thrombotic microangiopathy but also has distinctive features (Levi et al., 2020). Given the significance of CAC to COVID-19 mortality, there is an urgent need for fine-grained resolution into the temporal manifestation of CAC, particularly in comparison to the broad-spectrum of other, better characterized coagulopathies. While there are studies suggesting associations between COVID-19 infection and mortality with thrombocytopenia, D-dimer levels, and prolongation of prothrombin time, the signatures of CAC onset and progression as well as their connection to clinical outcomes are not well defined (Tang et al., 2020; Gao et al., 2020; Panigada et al., 2020). An advanced understanding of this phenotype may aid in the risk stratification of patients, thus facilitating optimal monitoring strategies during disease evolution through the paradigm of precision medicine.

To this end, we instituted a holistic data science platform across an academic medical center that enables machine intelligence to augment the curation of phenotypes and outcomes from over 10 million electronic health record (EHR) clinical notes and associated 3.2 million lab tests from 2232 SARS-CoV-2 positive (COVIDpos) and 72,354 confirmed SARS-CoV-2 negative (COVIDneg) patients over a retrospectively defined 2-month observation period straddling the date of the PCR test. For the COVIDpos cohort, we center the 2-month observation period around the date of the first positive PCR test for SARS-CoV-2, and for the COVIDneg cohort, we center the 2-month observation period around the date of the first PCR test for SARS-CoV-2 (see Materials and methods). It is important to note that not all individuals infected by SARS-CoV-2 develop symptoms of COVID-19, but rather that a majority of patients are either asymptomatic or have mild-to-moderate symptoms not requiring hospitalization for COVID-19 (Wagner et al., 2020). Furthermore, the guidelines followed for PCR-testing included a routine screening of individuals, patients displaying COVID-19 symptoms as per the Mayo Clinic (Coronavirus disease, 2019) and CDC definitions (Website, 2020), and possibly contact with infected persons or underlying predisposing conditions (Wagner et al., 2020).

By compiling all available laboratory testing data for the 30 days preceding the first SARS-CoV-2 PCR positive diagnostic testing date (day 0), as well as the 30 days following the diagnostic testing date, and triangulating this information with medications and clinical outcomes, we were able to identify laboratory abnormalities significantly associated with the COVIDpos group. We identified coagulation-related parameters among this set of abnormalities and then studied aggregate as well as individual patient trajectories that could aid in extracting a temporal signature of CAC onset and progression. We also correlated these signals with the clinical outcomes of these patients.

In order to hone into longitudinal lab test trends that would apply at the individual patient level, we restricted our analysis to patients with available serial testing data, which had at least three test results of the same type during the observation period. After applying these inclusion criteria, 246 COVIDpos and 13,666 COVIDneg patients met study the inclusion criteria. The need for longitudinal data on the testing results, while constraining the study population size greatly, enables us to provide a fine-grained temporal resolution of CAC for the first time.

After filtering the patients with the available longitudinal testing data, the median age in the COVIDpos and COVIDneg groups were 60.8 years and 64.1 years, respectively (see Materials and methods and Table 1), and the numbers of males were 137 (56%) and 7129 (52%), respectively. The total numbers of pre-existing coagulopathies in the COVIDpos and COVIDneg groups were 31 (13%) and 3901 (29%), respectively. These counts of coagulopathies include the following phenotypes identified in the clinical notes from day −365 to day −31 relative to the PCR testing date: deep vein thrombosis, pulmonary embolism, myocardial infarction, venous thromboembolism, thrombotic stroke, cerebral venous thrombosis, and disseminated intravascular coagulation (see Table 1 for detailed breakdown). The number of COVIDpos patients hospitalized in the month prior to the SARS-CoV-2 PCR testing date was 41 (17%), compared to 1247 (9.1%) for the COVIDneg cohort.

Table 1
Summary of patient characteristics for the overall COVIDpos, COVIDneg (matched), and COVIDneg cohorts.

The COVIDneg (matched) cohort was constructed using 1:10 propensity score matching to balance each of the clinical covariates, including demographics (age, gender, race), medication use (anticoagulant/antiplatelet use in the preceding 30 days/1 year of PCR testing date), medical history of thrombotic events from the past year, and hospitalization status in the month prior to the date of PCR testing.

Patient characteristicsCOVIDposCOVIDneg (matched)COVIDneg
Number of patients246246013,666
Age in years60.860.964.1
Gender:
Male137 (56%)1388 (56%)7129 (52%)
Race:
White154 (63%)1540 (63%)12,241 (90%)
Black24 (9.8%)313 (13%)569 (4.2%)
Asian18 (7.3%)207 (8.4%)274 (2.0%)
American Indian23 (9.3%)81 (3.3%)81 (0.59%)
Other27 (11%)319 (13%)501 (3.7%)
Medication use in the preceding 30 days of PCR testing date:
Anticoagulants63 (26%)596 (24%)5171 (38%)
Antiplatelets30 (12%)298 (12%)2230 (16%)
Medication use in the preceding 1 year of PCR testing date:
Anticoagulants86 (35%)819 (33%)7476 (55%)
Antiplatelets40 (16%)419 (17%)3620 (26%)
Medical history of thrombotic events in 1 year prior to study period:
Deep vein thrombosis15 (6.1%)153 (6.2%)2,110 (15%)
Pulmonary embolism12 (4.9%)112 (4.6%)1258 (9.2%)
Myocardial infarction11 (4.5%)142 (5.8%)1468 (11%)
Venous thromboembolism4 (1.6%)44 (1.8%)615 (4.5%)
Thrombotic stroke1 (0.41%)3 (0.12%)143 (1.0%)
Cerebral venous thrombosis01 (0.04%)7 (0.05%)
Disseminated intravascular coagulation01 (0.04%)30 (0.22%)
Any thrombotic event31 (13%)308 (13%)3901 (29%)
Hospitalized in the month prior to PCR testing date41 (17%)304 (12%)1247 (9%)

To balance these clinical covariates and others between the two cohorts, we applied 1:10 propensity score matching to define a subset of 2460 patients from the COVIDneg cohort to use for the final statistical analysis (see Materials and methods). In particular, the general categories of covariates considered for balancing included: demographics, anticoagulant/antiplatelet medication use, medical history of pre-existing coagulopathies, and hospital admission status. Population-level characteristics of the COVIDpos, COVIDneg, and the final propensity score-matched COVIDneg (matched) cohorts are summarized in Table 1. We observe that the COVIDpos and COVIDneg (matched) cohorts are well-balanced along these covariates which are potential confounding variables for thrombotic events and coagulopathy-related lab tests during the study period.

Results

Longitudinal analysis identifies lab test results characteristic of COVID-19 at specific prognostic time intervals

To identify laboratory test results that differ between COVIDpos and COVIDneg (matched) patients, we analyzed longitudinal trends of 194 laboratory test results in the 30 days before and after the day of PCR testing (designated as day 0). As most patients did not undergo laboratory testing for each assay on a daily basis, we grouped the measurements into nine time windows reflecting potential stages of infection as follows: pre-infection (days −30 to −11), pre-PCR (days −10 to −2), time of clinical presentation (days −1 to 0), and post-PCR phases 1 (days 1 to 3), 2 (days 4 to 6), 3 (days 7 to 9), 4 (days 10 to 12), 5 (days 13 to 15), and 6 (days 16 to 30). We only considered test-time window pairs in which at least three patients contributing to laboratory test results in both groups. During each time window, we then compared the distribution of results from COVIDpos versus COVIDneg (matched) patients, allowing us to identify any lab tests which were significantly altered in COVIDpos patients during any time of disease acquisition, onset, and/or progression.

Of the 1709 lab test-time window pairs with adequate data points for comparison, we identified 130 such pairs (comprising 66 unique lab tests) which met our thresholds for statistical significance (Cohen’s D >0.35, BH-adjusted Mann-Whitney p-value <0.05; Table 2). Among these were lab tests that may be considered positive controls for our analysis. From the time of clinical presentation onward, elevated titers of SARS-CoV-2 IgG antibodies (Figure 1A) and a reduction in blood oxygenation in COVIDpos patients were observed (Figure 1B). We also identified abnormalities in several other classes of lab tests, including immune cell counts (Figures 1C–E and 2A–B), red blood cell counts (Figure 2C), mean corpuscular volume (Figure 2D), calcium and magnesium levels (Figure 2E–F), and coagulation-related tests (Figure 3).

Longitudinal and temporally resolved analysis highlights positive control lab tests elevated in COVIDpos patients along with distinctive immune signatures.

Longitudinal trends in COVIDpos versus COVIDneg (matched) patients for the following lab tests: (A) SARS-CoV-2 IGG ratio, (B) oxygen saturation in arterial blood, (C) white blood cells, (D) monocytes absolute, and (E) neutrophils, blood. For any window of time during which at least three patients in each cohort had test results, data are shown as mean with standard errors. The normal range for each lab test is shaded in green. Values given horizontally along the top of the plot are Cohen’s D statistics comparing the COVIDpos and COVIDneg (matched) cohorts along with the BH-adjusted Mann-Whitney test p-values. Significant differences (adjusted p-value <0.05) are shown in black, while non-significant values are shown in gray. Values given horizontally along the bottom of the plot are the numbers of patients in the COVIDpos and COVIDneg cohorts, respectively (i.e. # COVIDpos | # COVIDneg). For certain lab tests, some data points are missing because these time windows had fewer than three data points in the COVIDpos cohort.

Longitudinal trends of COVIDpos patients’ lab tests show distinctive immune, hematologic, and serum chemistry signatures within normal ranges.

Longitudinal trends in COVIDpos versus COVIDneg (matched) patients for the following lab tests: (A) eosinophils absolute, (B) basophils absolute, (C) red blood cell count, (D) mean corpuscular volume, (E) calcium total, plasma, and (F) magnesium total, serum/plasma. For any window of time during which at least three patients in each cohort had test results, data are shown as mean with standard errors. The normal range for each lab test is shaded in green. Values given horizontally along the top of the plot are Cohen’s D statistics comparing the COVIDpos and COVIDneg (matched) cohorts along with the BH-adjusted Mann-Whitney test p-values. Significant differences (adjusted p-value <0.05) are shown in black, while non-significant values are shown in gray. Values given horizontally along the bottom of the plot are the numbers of patients in the COVIDpos and COVIDneg cohorts, respectively (i.e. # COVIDpos | # COVIDneg). For certain lab tests, some data points are missing because these time windows had fewer than three data points in the COVIDpos cohort.

COVIDpos patients show distinctly opposite temporal trends in fibrinogen and platelet counts starting at the time of diagnosis.

Longitudinal trends of COVIDpos versus COVIDneg (matched) patients for the following lab tests: (A) fibrinogen, plasma, (B) platelets, and (C) other coagulation-related tests including prothrombin time (PT), activated partial thromboplastic time (aPTT), and D-dimers. For any window of time during which at least three patients in each cohort had test results, data are shown as mean with standard errors. The normal range for each lab test is shaded in green. Values given horizontally along the top of the plot are Cohen’s D statistics comparing the COVIDpos and COVIDneg (matched) cohorts along with the BH-adjusted Mann-Whitney test p-values. Significant differences (adjusted p-value <0.05) are shown in black, while non-significant values are shown in gray. Values given horizontally along the bottom of the plot are the numbers of patients in the COVIDpos and COVIDneg cohorts, respectively (i.e. # COVIDpos | # COVIDneg). For certain lab tests, some data points are missing because these time windows had fewer than three data points in the COVIDpos cohort.

Table 2
Summary of lab tests significantly different between COVIDpos and propensity score-matched COVIDneg cohorts during at least one clinical time window.

Data from individual patients were averaged over the defined time windows, and the mean values were compared between COVIDpos and COVIDneg patients. The lab test-time window pairs shown are those which met our defined thresholds for statistical significance and substantial effect (BH-adjusted Mann-Whitney p-value <0.05 and Cohen’s D absolute value >0.35). In particular, 130 of the initial 1709 (test, time window) pairs with at least one patient met these thresholds. Rows are sorted alphabetically by test and then time window (from earliest to latest). Coagulation-related tests of particular interest (fibrinogen, platelets, prothrombin time, activated partial thromboplastin time, and D-dimer) are highlighted in gray. Sample sources are denoted as: P = plasma, S = serum, S/P = serum/plasma, B = blood, U = urine.

TestUnitsTime windowCount COVIDposCount COVIDnegMean COVIDposMean COVIDnegCohen's DBH-adj M-W p-value
ABGRS pH ArterialpHDays 16–30 Post-Dx18917.457.40.7750.02
ABGRS PO2 Arterialmm HgDays 1–3 Post-Dx1620481.9129.6−0.7973.1E-03
ABGRS PO2 Arterialmm HgDays 4–6 Post-Dx258278.1113.2−0.7128.8E-03
ABGRS PO2 Arterialmm HgDays 7–9 Post-Dx235877.2121.9−0.8071.0E-03
ABGRS PO2 Arterialmm HgDays 10–12 Post-Dx183776.4104.2−0.9652.6E-03
ABGRS PO2 Arterialmm HgDays 13–15 Post-Dx153173.1112.3−0.9646.0E-03
Activated Partial Thrombopl Time, PsecDays 7–9 Post-Dx226650.536.70.7270.026
Activated Partial Thrombopl Time, PsecDays 10–12 Post-Dx145463.339.21.0852.4E-03
Activated Partial Thrombopl Time, PsecDays 13–15 Post-Dx164853.137.61.0655.6E-03
Activated Partial Thrombopl Time, PsecDays 16–30 Post-Dx1914956.237.50.8840.027
Alanine Aminotransferase (ALT), PU/LDays 10–12 Post-Dx2710477.3460.5120.015
Albumin, Pg/dLDays 7–9 Post-Dx421883.063.41−0.545.6E-03
Albumin, S/Pg/dLClinical presentation858123.433.81−0.5914.8E-06
Albumin, S/Pg/dLDays 1–3 Post-Dx775253.263.6−0.5413.8E-05
Albumin, S/Pg/dLDays 10–12 Post-Dx612543.353.66−0.472.6E-03
Alkaline Phosphatase, PU/LDays 4–6 Post-Dx4213988.8126.7−0.3953.7E-03
Arterial O2 PP DiffNoneClinical presentation21106268.1152.10.9249.7E-03
Arterial O2 PP DiffNoneDays 1–3 Post-Dx22112225.9147.40.6390.017
Arterial O2 PP DiffNoneDays 4–6 Post-Dx1749271.41550.8914.8E-03
Aspartate Aminotransferase (AST), PU/LDays 10–12 Post-Dx2710767.644.70.4043.6E-04
Basophils Absolute×10(9)/LClinical presentation13314000.02510.0379−0.4125.8E-06
Bicarbonate [MMOL/L] in Arterial Bloodmmol/LDays 16–30 Post-Dx189128.624.30.8577.6E-03
Bicarbonate in Arterial Bloodmmol/LDays 1–3 Post-Dx2619323.221.40.5130.027
BUN, Pmg/dLDays 16–30 Post-Dx4956231.421.90.5553.9E-03
C-reactive Protein Quantative, Smg/LClinical presentation85666100.268.20.3756.8E-05
Calcium, Ionized, Bmg/dLClinical presentation142014.364.77−0.670.015
Calcium, Ionized, Bmg/dLDays 1–3 Post-Dx182704.424.73−0.7838.5E-04
Calcium, Total, Pmg/dLClinical presentation8911448.719.05−0.4685.5E-06
Calcium, Total, Pmg/dLDays 1–3 Post-Dx779108.528.81−0.4593.2E-04
Calcium, Total, Pmg/dLDays 7–9 Post-Dx713538.618.93−0.4571.8E-03
Calcium, Total, Smg/dLClinical presentation839418.298.91−0.8541.9E-13
Calcium, Total, Smg/dLDays 1–3 Post-Dx9810258.288.77−0.7172.2E-10
Calcium, Total, Smg/dLDays 4–6 Post-Dx875688.48.69−0.4352.3E-03
Calcium, Total, Smg/dLDays 7–9 Post-Dx824338.498.76−0.3840.011
Carboxyhemoglobin, ARTERIAL%Clinical presentation343560.5070.991−0.712.0E-04
Carboxyhemoglobin, Arterial%Days 1–3 Post-Dx444360.5350.9−0.7115.9E-05
Carboxyhemoglobin, Arterial%Days 4–6 Post-Dx581660.6780.974−0.5443.0E-03
Carboxyhemoglobin, Arterial%Days 7–9 Post-Dx451020.7040.97−0.4720.048
Carboxyhemoglobin, Venous%Days 1–3 Post-Dx10730.7011.16−0.8620.02
Carboxyhemoglobin, Venous%Days 4–6 Post-Dx14470.7251.29−0.8373.7E-03
Chloride, Pmmol/LDays 1–3 Post-Dx77906100.1101.9−0.3637.7E-03
Eosinophils Absolute×10(9)/LPre-diagnosis285470.06890.161−0.451.7E-03
Esosinophils Absolute×10(9)/LDays 4–6 Post-Dx1335590.09060.172−0.3582.4E-06
Fibrinogen, Pmg/dLClinical presentation51233528.9360.70.8598.9E-07
Fibrinogen, Pmg/dLDays 1–3 Post-Dx18319432.6297.40.8361.7E-03
Fibrinogen, Pmg/dLDays 4–6 Post-Dx26116477.8333.70.7440.014
Glucose, Random, Smg/dLDays 13–15 Post-Dx49314150126.50.5440.013
Hematocrit, B%Days 1–3 Post-Dx158158236.533.80.4339.6E-06
Hematocrit, B%Days 4–6 Post-Dx1528513632.10.6212.2E-10
Hematocrit, B%Days 7–9 Post-Dx13263935.531.80.5875.8E-08
Hematocrit, B%Days 10–12 Post-Dx11050535.131.80.5111.7E-05
Hemoglobin Arterialg/dLDays 1–3 Post-Dx3120812.110.80.6510.025
Hemoglobin, Bg/dLDays 1–3 Post-Dx158168211.911.10.3582.2E-04
Hemoglobin, Bg/dLDays 4–6 Post-Dx15287311.810.40.6361.4E-10
Hemoglobin, Bg/dLDays 7–9 Post-Dx13265311.610.40.562.0E-07
Hemoglobin, Bg/dLDays 10–12 Post-Dx11051611.410.30.492.6E-05
Ionized Calcium, Arterialmg/dLDays 16–30 Post-Dx8364.934.481.5610.022
Lactate Dehydrogenase, SU/LDays 10–12 Post-Dx2188406.2295.20.4631.4E-03
Lactate, Pmmol/LClinical presentation899541.371.93−0.4623.1E-06
Lymphocytes Percent%Days 13–15 Post-Dx56633.2151.5140.048
Lymphs Absolute×10(9)/LDays 13–15 Post-Dx563493.121.110.440.018
Magnesium, Plasmamg/dLDays 10–12 Post-Dx20872.141.910.7720.015
Magnesium, S/Pmg/dLDays 4–6 Post-Dx472792.221.980.7433.0E-03
Magnesium, S/Pmg/dLDays 7–9 Post-Dx402152.311.971.064.1E-06
Magnesium, S/Pmg/dLDays 10–12 Post-Dx361872.261.911.0052.9E-06
Magnesium, S/Pmg/dLDays 13–15 Post-Dx351792.221.890.9041.8E-07
Magnesium, S/Pmg/dLDays 16–30 Post-Dx333172.131.890.9061.6E-04
Manual Diff Promyelocytes%Days 1–3 Post-Dx6550.2501.4020.027
Mean Corpuscular VolumefLDays 10–12 Post-Dx11050289.592−0.388.8E-03
Methemoglobin, ABG%Clinical presentation343560.3350.571−0.6296.0E-03
Methemoglobin, ABG%Days 1–3 Post-Dx444360.4250.697−0.4631.5E-03
Monocytes Absolute×10(9)/LDays 1–3 Post-Dx13110790.4470.748−0.5022.6E-16
Monocytes Absolute×10(9)/LDays 4–6 Post-Dx1355840.4750.715−0.5972.2E-10
N-terminal-PRO-Brain Type Natriuretic Peptide, Spg/mLDays 4–6 Post-Dx1063415.67609.7−0.5252.9E-03
Neutrophils, B×10(9)/LClinical presentation13613825.317.12−0.3966.3E-06
Neutrophils, B×10(9)/LDays 1–3 Post-Dx13011414.736.32−0.3855.8E-05
NT-PRO BNP, Ppg/mLClinical presentation253721372.45327.9−0.3850.046
NT-PRO BNP, Ppg/mLDays 4–6 Post-Dx1420815.34388.8−0.9290.02
Nucleated RBC/100 WBCDays 13–15 Post-Dx231891.240.4470.5611.7E-03
OHB%Days 1–3 Post-Dx1324288.695−1.372.2E-03
OHB%Days 4–6 Post-Dx329092.193.7−0.3560.013
OHB%Days 7–9 Post-Dx244691.594.5−0.7013.3E-04
Osmolality, UmOsm/kgPre-diagnosis480231.5478.8−1.5090.044
Oxygen Content, Arterialvol %Days 4–6 Post-Dx32891613.70.8392.4E-03
Oxygen Saturation (%) in Arterial Blood%Clinical presentation2718994.296.2−0.523.1E-03
Oxygen Saturation (%) in Arterial Blood%Days 1–3 Post-Dx3121694.397.1−1.2938.4E-09
Oxygen Saturation (%) in Arterial Blood%Days 4–6 Post-Dx267094.395.7−0.5780.014
Oxygen Saturation (%) in Arterial Blood%Days 10–12 Post-Dx182993.496.5−1.2543.1E-03
Oxygen Saturation (%) in Arterial Blood%Days 13–15 Post-Dx172894.896.4−0.6710.043
pH Blood ArterialNoneDays 1–3 Post-Dx261937.427.390.5390.035
pH Blood VenouspHDays 1–3 Post-Dx10827.427.360.9630.031
pH, POCT, BNoneClinical presentation132027.417.330.7080.042
Platelets×10(9)/LPre-diagnosis39649184.8225.9−0.3930.024
PO2mm HgDays 1–3 Post-Dx814567.2179.7−1.3011.7E-03
PO2mm HgDays 7–9 Post-Dx141671.1121.1−0.9490.027
PO2 Arterialmm HgDays 1–3 Post-Dx26193100.4150.9−0.878.2E-05
PO2 Arterialmm HgDays 10–12 Post-Dx172593.6134−0.7550.019
Potassium, Smmol/LPre-diagnosis103983.934.35−0.8360.049
RABG Calculated O2 Hemoglobin%Days 1–3 Post-Dx2210993.695−0.4642.9E-03
RABG Calculated O2 Hemoglobin%Days 4–6 Post-Dx164993.295.3−0.8592.3E-03
RABG Calculated O2 Hemoglobin%Days 10–12 Post-Dx13229496.3−1.2690.038
RABG PF RatioNoneDays 4–6 Post-Dx17491.462.68−1.4896.9E-05
RABG PF RatioNoneDays 7–9 Post-Dx13221.752.56−1.0060.038
RABG PF RatioNoneDays 10–12 Post-Dx13221.833.22−1.5183.9E-03
RBC (Red Blood Cell) Count×10(12)/LClinical presentation15116714.323.990.4092.0E-04
RBC (Red Blood Cell) Count×10(12)/LDays 1–3 Post-Dx15815624.133.730.5245.8E-08
RBC (Red Blood Cell) Count×10(12)/LDays 4–6 Post-Dx1528464.083.550.6933.2E-12
RBC (Red Blood Cell) Count×10(12)/LDays 7–9 Post-Dx13263543.490.6562.4E-09
RBC (Red Blood Cell) Count×10(12)/LDays 10–12 Post-Dx1105023.953.480.5876.1E-07
Red Cell Distribution Width CV%Days 4–6 Post-Dx13772214.115.1−0.3733.4E-04
Red Cell Distribution Width CV%Days 7–9 Post-Dx11955214.215.4−0.4319.8E-05
Red Cell Distribution Width CV%Days 10–12 Post-Dx9742914.515.7−0.3941.2E-03
Sodium, Pmmol/LClinical presentation891141135.6137.3−0.3757.3E-03
Sodium, Pmmol/LDays 1–3 Post-Dx77927136.6138.1−0.3774.7E-03
Sodium, Smmol/LDays 10–12 Post-Dx69334140.8138.30.6512.0E-04
Spont. Breaths/minNoneDays 4–6 Post-Dx23672520.20.7670.016
Tacrolimus, Bng/mLDays 7–9 Post-Dx8814.228.12−1.1028.8E-03
Tacrolimus, Bng/mLDays 10–12 Post-Dx8793.89.24−1.4682.5E-03
Tacrolimus, Bng/mLDays 13–15 Post-Dx7713.78.52−1.477.5E-03
Tacrolimus, Bng/mLDays 16–30 Post-Dx101104.937.8−1.0940.022
TemperatureNoneClinical presentation231363736.70.5910.042
TemperatureNoneDays 1–3 Post-Dx231893736.40.7654.8E-04
Triglycerides, S/Pmg/dLDays 4–6 Post-Dx1641326.21731.1967.3E-03
Triglycerides, S/Pmg/dLDays 7–9 Post-Dx1724310.6191.50.9450.016
Triglycerides, S/Pmg/dLDays 10–12 Post-Dx1735364.5174.41.2174.0E-03
Triglycerides, S/Pmg/dLDays 16–30 Post-Dx1077276.1166.40.830.024
Troponin T, 5TH GEN, Png/LDays 4–6 Post-Dx185421.4245.3−0.4997.5E-03
Troponin T, Baseline, 5TH Gen, Png/LDays 7–9 Post-Dx114315.153.7−0.5380.037
VBGRS HGBg/dLDays 4–6 Post-Dx369912.310.50.9323.6E-04
White Blood Cells×10(9)/LDays 1–3 Post-Dx15816506.679.08−0.4393.2E-12

With respect to coagulation, we found that plasma fibrinogen was significantly elevated in COVIDpos patients at the time of diagnosis (Cohen’s D = 0.859, BH-adjusted Mann-Whitney p-value = 8.9e-7, Table 2, Figure 3A). This hyperfibrinogenemia generally resolved during the 7 days following diagnosis (Figure 3A). Conversely, platelet counts were lower in the COVIDpos cohort at the time of clinical presentation but tended to increase over the subsequent 10 days to levels significantly higher than those in COVIDneg patients (Cohen’s D = 0.229, BH-adjusted Mann-Whitney p-value = 3.6e-3, Table 2, Figure 3B). While thrombocytopenia has been reported in COVID-19 patients before (Xu et al., 2020; Yang et al., 2020), an upward trend in platelet counts after diagnosis has not been described to our knowledge. We observe extended prothrombin times in both the COVIDpos and COVIDneg (matched) cohorts significantly above the normal range; however, there was no differentiation between the cohorts. We observe extended activated partial thromboplastin times (aPTT) in the COVIDpos significantly above normal levels from day 7 onward (Figure 3D). D-dimer levels were frequently above normal limits in both the COVIDpos and COVIDneg cohorts and were not significantly different between these cohorts during any time window (Figure 3E). The above trends hold up even when the time windows are perturbed (Table 3).

Table 3
Sensitivity analysis of clinical time intervals for significant coagulation-related lab test trends.

Results from sensitivity analysis perturbing the time intervals for the significant (coagulation-related lab test, time interval) pairs (i.e. highlighted rows of Table 2). Perturbed results that met both of the significance thresholds (BH-adjusted Mann-Whitney p-value <0.05 and Cohen’s D absolute value >0.35) are highlighted in light green, and perturbed results that only met one of the thresholds for either effect size or statistical significance are highlighted in yellow.

TestUnitsPerturbationOriginal time windowCount COVIDposCount COVIDnegMean COVIDposMean COVIDnegCohen's DBH-adjusted
M-W p-value
Activated Partial Thrombopl Time, Psec−1 dayDays 7−9 Post-Dx267250.1380.570.034
Activated Partial Thrombopl Time, Psec+1 dayDays 7−9 Post-Dx17585537.50.810.014
Activated Partial Thrombopl Time, Psec−1 dayDays 10−12 Post-Dx165756.938.40.8089.10E-03
Activated Partial Thrombopl Time, Psec+1 dayDays 10−12 Post-Dx156056.9381.1062.60E-03
Activated Partial Thrombopl Time, Psec−1 dayDays 13−15 Post-Dx155255.537.81.0410.014
Activated Partial Thrombopl TIME, Psec+1 dayDays 13−15 Post-Dx144851.837.10.9620.015
Activated Partial Thrombopl Time, Psec−1 dayDays 16−30 Post-Dx2215655.2370.9135.70E-03
Activated Partial Thrombopl Time, Psec+1 dayDays 16−30 Post-Dx191395638.20.7253.80E-02
Fibrinogen, Pmg/dL−1 dayClinical presentation2592584.9370.71.0671.20E-04
Fibrinogen, Pmg/dL+1 dayClinical presentation37292488.2326.20.8858.80E-06
Fibrinogen, Pmg/dL−1 dayDays 1−3 Post-Dx41381494.53181.0233.90E-07
Fibrinogen, Pmg/dL+1 dayDays 1−3 Post-Dx21244420.3312.20.6167.90E-03
Fibrinogen, Pmg/dL−1 dayDays 4−6 Post-Dx27156432.23360.4950.045
Fibrinogen, Pmg/dL+1 dayDays 4−6 Post-Dx24105472.2333.20.7120.025
Plateletsx10(9)/L−1 dayPre-diagnosis34575187.3225.6-0.3570.057
Plateletsx10(9)/L+1 dayPre-diagnosis1181533201.3234.4-0.3287.30E-04

We also performed similar analyses comparing the COVIDpos and COVIDneg (matched) cohorts using different time window definitions including daily trends (Figure 4). This approach offers the advantage of increased granularity at the cost of sample size per time point, but we did identify similar lab tests as altered in COVIDpos patients using each approach including the fibrinogen decline and platelet increase in the COVIDpos cohort after diagnosis (Figure 4).

Longitudinal trends of lab tests with daily resolution.

Longitudinal trends of COVIDpos versus COVIDneg (matched) patients for the following lab tests: (A) platelets; (B) fibrinogen, plasma; (C) prothrombin time, plasma; (D) activated partial thromboplastin time; (E) D-dimer; (F) magnesium, serum/plasma; (G) basophils absolute; (H) neutrophils, blood; (I) alkaline phosphatase, serum. The reference ranges are shown at the top of each plot. For each cohort, average lab values and standard errors are shown for each day with at least three observations. For certain lab tests, some data points are missing because these days had fewer than three data points in the COVIDpos cohort.

Thrombosis is enriched among COVID-19 patients undergoing longitudinal lab testing

Given the recently described coagulopathies associated with COVID-19 (Tang et al., 2020; Klok et al., 2020; Levi et al., 2020), we were intrigued by the temporal trends in fibrinogen levels and platelet counts in the COVIDpos cohort (Figure 3). Next, we asked whether the observed coagulation-related laboratory trends were associated with clinical manifestations of thrombosis. To do so, we employed a BERT-based neural network (Devlin et al., 2018; see Materials and methods) to identify patients who experienced a thrombotic event after their SARS-CoV-2 PCR testing date. Specifically, we extracted diagnostic sentiment from EHR notes (e.g. whether a patient was diagnosed with a phenotype, suspected of having a phenotype, ruled out for having a phenotype, or other) regarding specific thromboembolic phenotypes including deep vein thrombosis, pulmonary embolism, myocardial infarction, venous thromboembolism, thrombotic stroke, cerebral venous thrombosis, and disseminated intravascular coagulation.

We found that 101 of the total 2232 COVIDpos cohort (4.5%) were positively diagnosed with one or more of the above-mentioned thrombotic phenotypes in the 30 days after PCR testing, with the majority of these patients (53 of 101) experiencing a deep vein thrombosis. Interestingly, we found that after creating subsets of the patients with longitudinal lab testing data (i.e. the patients meeting the criteria for inclusion in our study), 76 of the 246 patients (31%) had at least one EHR-derived clot diagnosis, including 47 patients with deep vein thrombosis (Table 4). Thus, the cohort under consideration here is highly enriched (Table 5; hypergeometric p-value <1×10−50) for patients experiencing thrombotic events compared to the overall COVIDpos cohort.

Table 4
Prevalence of thrombotic phenotypes after the clinical presentation in COVIDpos patients with and without available longitudinal lab testing data.

For each clotting phenotype listed, a BERT-based neural network was used to extract diagnostic sentiment from individual EHR patient notes in which the phenotype (or a synonym thereof) was present. This automated curation was applied to clinical notes for each patient from day = −1 (clinical presentation) to day = 30 (end of the study period) relative to the PCR testing date. In this table, we show the absolute number of patients with each phenotype along with the percentage of patients in each cohort with the given specific thrombotic phenotype in parentheses.

Clotting phenotypeCohort 1: COVIDpos with longitudinal dataCohort 2: COVIDpos without longitudinal dataCohort 3: Complete COVIDpos cohort
Deep vein thrombosis47 (19%)6 (0.30%)53 (2.4%)
Pulmonary embolism22 (8.9%)9 (0.45%)31 (1.4%)
Myocardial infarction10 (4.1%)8 (0.40%)18 (0.81%)
Venous thromboembolism7 (2.8%)07 (0.31%)
Thrombotic stroke2 (0.81%)2 (0.10%]4 (0.18%)
Cerebral venous thrombosis000
Disseminated intravascular coagulation5 (2.0%)05 (0.22%)
Total unique patients with clot76 (31%)25 (1.3%)101 (4.5%)
Total patients24619862232
Table 5
Enrichment of thrombotic phenotypes among COVIDpos patients with longitudinal lab testing data.

Contingency table to calculate hypergeometric enrichment significance of thrombosis among patients with longitudinal lab testing data. The 246 patients with longitudinal testing data are those considered in this study, while the 1986 patients who did not have at least three results from one lab test over the defined 60-day window were excluded from this longitudinal analysis.

Patient has longitudinal dataPatient does NOT have longitudinal dataTotal
Thrombosis7625101
No thrombosis17019612131
Total24619862232
  1. Hypergeometric enrichment: p-value <1×10−50.

Table 6
Validation of the BERT model to identify the sentiment of thrombotic phenotypes in clinical notes.

Out-of-sample accuracy results of the BERT model to identify thrombotic phenotypes in 1000 randomly selected sentences from clinical notes which contained at least one mention of a thrombotic phenotype. The columns are (1) Clotting phenotype: thrombotic phenotype identified in the sentence, (2) TP (true positives): count of sentences in which the BERT model correctly identified the sentiment as ‘Yes’, (3) TN (true negatives): count of sentences in which the BERT model correctly identified the sentiment as not ‘Yes’, (4) FP (false positives): count of sentences in which the BERT model incorrectly identified the sentiment as ‘Yes’, (5) FN: (false negatives): count of sentences in which the BERT model incorrectly identified the sentiment as not ‘Yes’, (6) Recall: recall of the BERT model, equal to TP/(TP+FN), (7) Precision: precision of the BERT model, equal to TP/(TP+FP), (8) Accuracy: accuracy of the BERT model, equal to (TP+TN)/(TP+TN+FP+FN).

Clotting phenotypeTPTNFPFNRecallPrecisionAccuracy
Deep vein thrombosis13617824398%85%92%
Pulmonary embolism164787696%96%95%
Myocardial infarction212653399%99%98%
Venous thromboembolism39770100%30%93%
Thrombotic stroke5000100%100%100%
Cerebral venous thrombosis1000100%100%100%
Disseminated intravascular coagulation4400100%100%100%
Overall525422411297.8%92.8%94.7%

Longitudinal platelet count trends are not strongly associated with the development of thrombosis in COVID-19 patients

Among the 246 COVIDpos patients with longitudinal lab testing data, 81 were serially tested starting at clinical presentation for fibrinogen versus 245 tested for platelets. As such, we first analyzed whether associations exist between platelet counts (or temporal alterations thereof) and clotting propensity in this cohort. Among these 245, there were 169 patients without thrombosis after PCR-based diagnosis (non-thrombotic) and 76 patients with thrombosis (thrombotic). There is a statistically significant difference between the COVIDpos and COVIDneg cohorts in the platelet count at clinical presentation (Figure 5A). In particular, thrombocytopenia (platelet count <150×109/L) was observed in 29% (46 out of 154) COVIDpos and 21% (346 of 1661) COVIDneg patients at the time of diagnosis (Figure 5A). However, the platelet levels at this time point were not associated with the subsequent formation of a blood clot in the COVIDpos cohort (Figure 5B).

Association between platelet counts and thrombosis in the COVIDpos cohort.

Box plots of platelet counts, min/max values, and maximum levels of increase/decline at specific time intervals for COVIDpos and COVIDneg cohorts and subgroups of the COVIDpos cohort with and without thrombotic events after SARS-CoV-2 diagnosis. In the subplot (A), we show platelet counts for COVIDpos (red) and COVIDneg (blue) cohorts. In subplots (B-F), we show platelet counts for COVIDpos patients who did and did not subsequently develop thromboses (purple and black, respectively). Horizontal dotted gray lines correspond to upper and lower limits of normal platelet counts (150−450 × 109/L), and horizontal red line shows 100 × 109/L. At the top of each plot, Cohen’s D effect size and p-value from the Mann-Whitney statistical test are shown. (A) Platelet counts at the time of PCR testing for COVIDpos and COVIDneg cohorts. (B) Platelet counts at the time of PCR testing for COVIDpos patients who did and did not subsequently develop thromboses. (C) Maximum platelet counts (considering counts at and after positive PCR test date) for COVIDpos patients who did and did not subsequently develop thromboses. (D) Minimum platelet counts (considering counts at and after positive PCR test date) for COVIDpos patients who did and did not subsequently develop thromboses. (E) Maximum degree of platelet increases after positive PCR test date for COVIDpos patients who did and did not subsequently develop thromboses. (F) Maximum degree of platelet declines after positive PCR test date for COVIDpos patients who did and did not subsequently develop thromboses.

We hypothesized that the previously discussed increase in platelet counts after COVID-19 diagnosis may be associated with the development of blood clots. If true, then we would expect the thrombotic COVIDpos cohort to show significantly higher maximum platelet counts during their course of disease progression compared to the non-thrombotic COVIDpos cohort. We found that this was not the case, as maximum platelet counts were similar in the two groups (Figure 5C). Similarly, among the 147 COVIDpos patients with platelet counts both at the time of clinical presentation and post-diagnosis, the degree of maximal platelet increase was not associated with the development of thrombosis (Figure 5E). It would certainly be of interest to perform this same analysis on a larger COVIDpos cohort (n = 2232; 101 thrombotic vs. 2131 non-thrombotic), but we were not able to do so given the lack of longitudinal testing available for a large majority of non-thrombotic COVIDpos patients (Table 4).

Conversely, we explored whether some COVIDpos patients may experience clotting in the setting of low or declining platelets (e.g. consumptive coagulopathy) despite the population-level trend of increasing platelets over time. Indeed, we found that nine of 74 thrombotic patients showed absolute platelet counts below 100 × 109/L during at least one post-diagnosis time window (below dotted red line in Figure 5D). In addition, we analyzed post-diagnosis platelet reductions among COVIDpos patients. While the maximum degree of absolute platelet reduction was not associated with clot development in aggregate (Figure 5F), we did find that six of the 52 thrombotic patients experienced a reduction of at least 100 × 109/L relative to the time of diagnosis. Of note, similar fractions of non-thrombotic COVIDpos patients also showed these low or declining platelet counts, indicating that these trends are not specific indicators of thrombosis (Figure 5D,F).

Consumptive coagulopathy contributes to only a small fraction of COVID-19 associated thromboses

The observed declining platelet counts and thrombocytopenia in the context of thrombosis in a small fraction of COVIDpos patients are consistent with previous reports that fewer than 1% of survivors, but over 70% of non-survivors, meet the International Society on Thrombosis and Hemostasis (ISTH) criteria for disseminated intravascular coagulation (DIC; Tang et al., 2020). As was previously noted, hyperfibrinogenemia was among the strongest lab test features distinguishing COVIDpos from COVIDneg patients at diagnosis, but the subsequent downward trend (Figure 3A) could be attributed to a resolving acute phase response and/or consumption of fibrinogen in a systemic coagulopathy. Using our BERT-based sentiment extraction, we found that only five of the 2232 COVIDpos patients that exhibited DIC-like symptoms, all of whom were included in our longitudinal cohort of 246 COVIDpos patients (Table 4). Upon manual review of the EHR data for each patient, we found that two out of these five patients had confirmed diagnosis of DIC, while the remaining had high clinical suspicion and pending tests for DIC. This finding suggests that declining fibrinogen after COVID-19 diagnosis typically represents a physiologic return to normal range rather than pathologic coagulation factor consumption. To further examine the plasma fibrinogen trends among COVID-19 patients with DIC, with non-DIC thrombosis, and without thrombosis, we examined patient-level lab test trends from 10 individuals who were tested for fibrinogen both at the time of diagnosis and at least two times subsequently. The 10 patients for individual analysis were selected as the first 10 individuals with longitudinal fibrinogen lab testing data available.

This patient-level analysis indeed revealed multiple distinct trajectories with respect to fibrinogen and other coagulation parameters in COVIDpos patients. Four of these ten individuals developed at least one blood clot during their hospital course. Only one was identified by our BERT model (and confirmed by manual EHR review) to have low-grade DIC, and as expected we found this patient’s longitudinal lab test pattern to be consistent with consumptive coagulopathy (Patient 124; Figure 6A). At the time of diagnosis, this patient showed significant hyperfibrinogenemia with elevated D-dimers (1304.5 ng/mL) and a borderline normal platelet count (153 × 109/L). Over the next 10 days, this patient’s fibrinogen levels consistently decreased, reaching a minimum of 110 mg/dL on day 9. Similarly, after an initial recovery to 190 × 109/L the platelet counts consistently declined starting on day 2 post-diagnosis, reaching a minimum of 117 × 109/L on day 11. D-dimer levels exponentially increased after 5 days, reaching a maximum of 41,300 ng/mL on day 10. Phenotypically, this patient experienced both thrombotic (right internal jugular vein and right superior thyroid artery) and hemorrhagic (oropharyngeal and pulmonary) events. This combination of lab results and clinical manifestations is consistent with the diagnosis of DIC-like consumptive coagulopathy during the first week after COVID-19 diagnosis.

Longitudinal analyses of platelet counts, plasma fibrinogen, and D-dimer levels in individual patients with or without thrombotic disease.

In each plot, shaded regions represent time periods when the patient was taking a specific anticoagulant or antiplatelet medication. Medications taken for prophylaxis are denoted in the legend with (ppx). (A) Patient 124 developed hemorrhagic and thrombotic phenotypes in the context of declining fibrinogen, declining platelets, and increasing D-dimers. This is consistent with a DIC-like coagulopathy. (B) Patient 23 developed clots in the setting of declining fibrinogen and elevated D-dimers but stable platelet counts which increased shortly thereafter. (C) Patient 79 developed clots while showing increases in platelet counts along with plasma fibrinogen and D-dimers. (D) Patient 94 developed clots with relatively stable platelet counts and steadily declining plasma fibrinogen. (E) Patient 13 did not develop clots or bleeding despite a coordinate decrease in platelet counts and fibrinogen which may be mistaken for a DIC-like coagulopathy. (F) Patient 51 did not develop clots despite showing a post-diagnosis decline in plasma fibrinogen similar to several patients in the thrombotic cohort.

Lab test results from three other non-DIC thrombotic patients with longitudinal fibrinogen testing confirm the presence of alternative forms of coagulopathy in the COVID-19 population. Patient 23 developed a clot on day 4 post-diagnosis in the context of a declining fibrinogen level and increasing D-dimers but steady platelet counts, which actually increased shortly thereafter (Figure 6B). Patient 79 developed several clots after day 3 post-diagnosis in the setting of upward trending platelets (which eventually exceed the upper limit of normal) and elevated levels of both fibrinogen and D-dimers (Figure 6C). Patient 94 developed a clot on day 8 post-diagnosis with relatively stable platelet counts within normal limits and steadily declining fibrinogen levels (Figure 6D).

One hypothesis is that early elevations in plasma fibrinogen contribute to the clotting observed in the non-DIC like COVIDpos cohort. This hypothesis may warrant further analysis in cohorts with more longitudinal fibrinogen data, but again it is important to note that several COVIDpos patients who presented with hyperfibrinogenemia did not go on to develop thromboses (Figure 6E–F). This emphasizes that a steady post-diagnosis decline in plasma fibrinogen may represent physiologic resolution of the acute phase response rather than a pathologic consumption of fibrinogen and other coagulation factors (Figure 6B,D–F).

Taken together, this analysis affirms that a DIC-like coagulopathy resulting in a combination of hemorrhage and thrombosis can develop in the setting of COVID-19 infection. However, the observations that DIC was formally diagnosed in only five of 2232 COVIDpos patients and emphasizes that consumptive coagulopathy is an exception rather than the rule as it pertains to thrombotic phenotypes in COVID-19 patients. These results should be considered as a preliminary characterization of COVID-associated coagulopathies (CAC) and will be updated as patient counts increase with the continued evolution of the COVID-19 pandemic.

Discussion

Many studies on clinical characteristics and lab tests are shedding light on the spectrum of hematological parameters associated with COVID-19 patients. In an initial study of 41 patients from Wuhan, the blood counts in COVIDpos patients showed leukopenia and lymphopenia, and prothrombin time and D-dimer levels were higher in ICU patients than in non-ICU patients (Huang et al., 2020). Another study based on 343 Wuhan COVIDpos patients found that a D‐dimer level of at least 2.0 µg/mL could predict mortality with a sensitivity of 92.3% and a specificity of 83.3% (Zhang et al., 2020). An independent study of 43 COVID-19 patients found significant differences between mild and severe cases in plasma interleukin‐6 (IL‐6), D‐dimers, glucose, thrombin time, fibrinogen, and C‐reactive protein (p<0.05; Gao et al., 2020). While such studies indeed highlight that hematological and inflammatory abnormalities are prevalent in COVIDpos, a high-resolution temporal understanding of how these parameters evolve in COVID-19 patients post diagnosis has not been established. Specifically, in the wake of accumulating evidence for hypercoagulability in COVIDpos patients, there are important clinical questions emerging regarding the necessity of and guidelines for thromboprophylaxis in patient management.

DIC-like consumptive coagulopathy in COVID-19 has been a point of concern in severely ill COVID-19 patients. Particularly in patients with ARDS, multiple organ dysfunction syndrome (MODS) is the predominant cause of death. A recent study suggested that DIC was associated with MODS during the early stage of ARDS and that persistent DIC may also have a role in this association (Gando et al., 2020). Our study focusing on COVID-19 patients with longitudinal lab data suggests that COVID-19 is indeed associated with modulation of coagulation related parameters such as platelet counts, fibrinogen levels, and clotting time (Figure 2). However, the majority of thrombotic events in COVID-19 patients with longitudinal lab testing are not the result of a DIC-like consumptive coagulopathy, as this only occurs in a small subset (Table 4).

The ability to derive this longitudinal understanding of COVID-19 progression, including laboratory abnormalities and their associated clinical manifestations, mandates the synthesis of structured and unstructured EHR data (e.g. lab tests and clinical notes) at a large scale. The fact that tens of thousands of patients have undergone SARS-CoV-2 testing at major academic medical centers (AMCs) provides an abundance of potential data to perform this analysis but also poses significant challenges from a practicality standpoint. Manual review and curation of patient trajectories and associated testing results is not practical. It is not likely to provide comprehensive or even entirely accurate individual patient records. Rather, triangulation across datasets, including lab measurements, clinical notes, and prescription information, using a scalable digitized approach to extract structured data along with sentiment-surrounded clinical phenotypes and outcomes enables us to efficiently perform this analysis in a timely fashion.

By developing and deploying such a digitized platform on the entirety of EHR data from a large AMC, we have identified in an unbiased manner, laboratory test-based abnormalities that differentiate COVIDpos patients from COVIDneg patients. The abnormalities in coagulation-related tests, including fibrinogen and platelets, were intriguing in the context of literature reporting the occurrence of various clotting phenotypes in COVID-19 patients, including DIC-like consumptive coagulopathies along with more isolated clotting events in the lungs, central nervous system, and other tissues (Tang et al., 2020; Klok et al., 2020; Levi et al., 2020). Our finding that consumptive coagulopathy represents a minority of COVID-19 associated clotting events provides context for other studies, which have reported overt DIC or DIC-like disease in over 70% of non-survivors but far lower fractions of survivors (Tang et al., 2020). As the pandemic continues to evolve and the patient counts increase over the coming months, we will be monitoring and reporting any updates to the clinical and laboratory observations drawn in this study.

Notwithstanding the preliminary nature of the analysis presented in this study, the results highlight that consumptive coagulopathy should be considered in the minority of COVIDpos patients with significant serial reductions in platelet counts. It remains to be seen whether the post-diagnosis platelet increases or early hyperfibrinogenemia which we observed may contribute mechanistically to the clotting in the much larger non-DIC thrombotic COVID-19 population. It is important to note that despite the trend of increasing platelets, the platelet count only extended above the normal range (>450×109/L) after the PCR date in few COVIDpos patients with serial measurements, and the development of such outright thrombocytosis was observed with similar frequencies in the thrombotic and non-thrombotic cohorts (Figure 5C). Further, the fact that several patients with elevated fibrinogen (i.e. >400 mg/dL) at presentation did not develop thromboses suggests that early hyperfibrinogenemia is not a singular driver of subsequent clotting events, but a small sample size (n = 10 patients; nine non-thrombotic vs. one thrombotic) limited the power of this analysis (Figure 6).

Despite these caveats, this linking of longitudinal trends to patient outcomes provides several useful pieces of clinical information. First, hyperfibrinogenemia is to be expected in COVID-19 patients around the time of diagnostic testing. Furthermore, declining fibrinogen levels shortly after diagnosis are also expected and likely represent the resolution of acute phase response in most patients rather than a decline secondary to the onset of consumptive coagulation. In addition, borderline or overt thrombocytopenia is common in COVID-19 patients at the time of clinical presentation, and the initial platelet count does not robustly predict patients who are likely to develop thromboses. After diagnosis, COVID-19 patients generally show an upward trend in platelets. Patients whose platelets trend down after diagnosis should be monitored, as platelet reductions after clinical presentation are associated with thromboses and significant reductions may be indicative of ongoing consumptive coagulopathy.

One unavoidable limitation of this study is that we restrict our analysis to patients which have longitudinal lab testing data available. While the inclusion criteria is naturally biased, we consider this study population to be of high clinical interest because these patients are highly enriched for severe thrombotic events during the study period (see Table 5). Further, in the propensity score matching step of the analysis, we are able to construct a control cohort that is similar to the COVIDpos cohort in these enriched dimensions. To provide additional color on the distinctive attributes of the study population, we provide a summary of the clinical characteristics of the study population versus all patients with PCR tests during the same time period (see Table 7). In addition, we provide the median numbers of lab tests per patient for selected coagulation-related lab tests (fibrinogen, platelets, PTT, APTT, D-dimer) and total lab tests (Tables 8 and 9).

Table 7
General characteristics of patients with SARS-CoV-2 PCR testing.

General demographic characteristics of all patients who underwent SARS-CoV-2 PCR testing in the Mayo Clinic EHR database from February 15, 2020 to May 28, 2020. Includes summary characteristics for: (A) all patients with at least one SARS-CoV-2 PCR test, and (B) patients with at least one SARS-CoV-2 PCR test and longitudinal testing data available (i.e. patient received the same lab test on 3 separate days within + / − 30 days of PCR testing date).

(A) Demographics of all patients with PCR testing data
COVIDposCOVIDneg
Total number of patients223272,354
Gender:
Male1153 (52%)31,613 (44%)
Female1074 (48%)40,714 (56%)
Race:
White1115 (50%)62,605 (87%)
Black420 (19%)2792 (3.9%)
Asian151 (6.8%)1719 (2.4%)
American Indian29 (1.3%)302 (0.42%)
Other517 (23%)4936 (6.8%)
(B) Demographics of patients with PCR testing data and longitudinal testing data
TestUnitsPerturbationOriginal time windowCount COVIDposCount COVIDnegMean COVIDposMean COVIDnegCohen's DBH-adjusted M-W p-value
Activated Partial Thrombopl Time, Psec−1 dayDays 7–9 Post-Dx267250.1380.570.034
Activated Partial Thrombopl Time, Psec+1 dayDays 7–9 Post-Dx17585537.50.810.014
Activated Partial Thrombopl Time, Psec−1 dayDays 10–12 Post-Dx165756.938.40.8089.10E-03
Activated Partial Thrombopl Time, Psec+1 dayDays 10–12 Post-Dx156056.9381.1062.60E-03
Activated Partial Thrombopl Time, Psec−1 dayDays 13–15 Post-Dx155255.537.81.0410.014
Activated Partial Thrombopl Time, Psec+1 dayDays 13–15 Post-Dx144851.837.10.9620.015
Activated Partial Thrombopl Time, Psec−1 dayDays 16–30 Post-Dx2215655.2370.9135.70E-03
Activated Partial Thrombopl Time, Psec+1 dayDays 16–30 Post-Dx191395638.20.7253.80E-02
Fibrinogen, Pmg/dL−1 dayClinical presentation2592584.9370.71.0671.20E-04
Fibrinogen, Pmg/dL+1 dayClinical presentation37292488.2326.20.8858.80E-06
Fibrinogen, Pmg/dL−1 dayDays 1–3 Post-Dx41381494.53181.0233.90E-07
Fibrinogen, Pmg/dL+1 dayDays 1–3 Post-Dx21244420.3312.20.6167.90E-03
Fibrinogen, Pmg/dL−1 dayDays 4–6 Post-Dx27156432.23360.4950.045
Fibrinogen, Pmg/dL+1 dayDays 4–6 Post-Dx24105472.2333.20.7120.025
Platelets×10(9)/L−1 dayPre-diagnosis34575187.3225.6−0.3570.057
Platelets×10(9)/L+1 dayPre-diagnosis1181533201.3234.4−0.3287.30E-04
Table 8
Lab test data availability in patients with SARS-CoV-2 PCR testing.

Lab test data availability for all patients who underwent SARS-CoV-2 PCR testing in the Mayo Clinic EHR database from February 15, 2020 to May 28, 2020. Includes counts of lab tests and counts of patients with 1+ and 3+ lab tests both overall and for selected coagulation-related lab tests (activated partial thromboplastin time, D-dimer, fibrinogen, platelets, and prothrombin time).

COVIDposCOVIDneg
Total number of patients223272,354
Number of patients with 1+ lab test566 (25%)35,188 (49%)
Number patents with 1+ test from day −30 to day −1299 (13%)23,116 (32%)
Number patents with 1+ test from day 0 to day 30452 (20%)28,666 (40%)
Number of patients with 3+ lab tests of the same type246 (11%)13,666 (19%)
Total number of lab tests98,75332,40,491
Number of lab tests from day −30 to day −112,12010,33,762
Number of lab tests from day 0 to day 3086,63322,06,729
ACTIVATED PTT
Number of lab tests3626042
Number of patients with 1+ lab test93 (4.0%)3544 (4.9%)
Number of patients with 3+ lab tests20 (0.86%)406 (0.56%)
D-DIMER, P
Number of lab tests9112846
Number of patients with 1+ lab test247 (11%)2395 (3.3%)
Number of patients with 3+ lab tests99 (4.4%)56 (0.077%)
FIBRINOGEN, P
Number of lab tests2783,017
Number of patients with 1+ lab test84 (3.8%)1217 (1.7%)
Number of patients with 3+ lab tests18 (0.81%)273 (0.38%)
PLATELETS
Number of lab tests26461,08,722
Number of patients with 1+ lab test500 (22%)30,732 (42%)
Number of patients with 3+ lab tests231 (10%)11544 (16%)
PROTHROMBIN TIME, P
Number of lab tests71128,007
Number of patients with 1+ lab test197 (8.8%)10,446 (14%)
Number of patients with 3+ lab tests46 (2.1%)2502 (3.5%)
Table 9
Lab test data availability in patients with SARS-CoV-2 PCR testing and longitudinal lab data.

Lab test data availability for all patients who underwent SARS-CoV-2 PCR testing in the Mayo Clinic EHR database from February 15, 2020 to May 28, 2020 with longitudinal testing data available (i.e. patient received the same lab test on three separate days within + / − 30 days of PCR testing date). Includes counts of lab tests and counts of patients with 1+ and 3+ lab tests both overall and for selected coagulation-related lab tests (activated partial thromboplastin time, D-dimer, fibrinogen, platelets, and prothrombin time).

COVIDposCOVIDneg
Total number of patients24613,666
Number patents with 1+ test from day −30 to day −1150 (61%)11,567 (85%)
Number patents with 1+ test from day 0 to day 30240 (98%)13,501 (99%)
Total number of lab tests89,5872,634,070
Number of lab tests from day −30 to day −18698763,808
Number of lab tests from day 0 to day 3080,8891,870,262
ACTIVATED PTT
Number of lab tests3555186
Number of patients with 1+ lab test86 (35%)2722 (20%)
Number of patients with 3+ lab tests20 (8.1%)406 (3.0%)
D-DIMER, P
Number of lab tests8551720
Number of patients with 1+ lab test197 (80%)1293 (9.5%)
Number of patients with 3+ lab tests99 (40%)56 (0.41%)
FIBRINOGEN, P
Number of lab tests2752965
Number of patients with 1+ lab test81 (33%)1168 (8.5%)
Number of patients with 3+ lab tests18 (7.3%)273 (2%)
PLATELETS
Number of lab tests234387,517
Number of patients with 1+ lab test245 (100%)13,399 (98%)
Number of patients with 3+ lab tests231 (94%)11,544 (84%)
PROTHROMBIN TIME, P
Number of lab tests67624,489
Number of patients with 1+ lab test165 (67%)7209 (53%)
Number of patients with 3+ lab tests46 (19%)2502 (18%)

It is important to note that while we center the study period around the PCR testing date, this date may not correspond to the same disease state of COVID-19 for each individual in the COVIDpos cohort. To account for the potential variability in disease progression, we have performed a sensitivity analysis on the time intervals (Table 3). Additionally, there are several covariates that may influence these longitudinal trends and should be explored further. For example, we have already considered whether previous or concomitant administration of anticoagulants or antiplatelet agents influences patient lab test results and/or outcomes. Similarly, in the future, we intend to explore whether longitudinal lab measurement trends differ between outpatient, inpatient, and ICU admitted patient cohorts. New datasets can also be utilized; for example, rather than grouping patients by the identified thromboembolic phenotypes extracted from the clinical notes alone, patients could be stratified by those who had imaging studies (duplex ultrasound, CT scan, etc.) performed, and phenotypes could be directly extracted from these procedural reports. As more data accumulates from COVIDpos and COVIDneg patients in the coming months, these analyses need to be expanded to assess similarities and differences in the temporal trends of laboratory test results among a wider range of patient subgroups relevant for COVID-19 outcomes, such as those who have pre-existing conditions (e.g. diabetes, hypertension, obesity, malignancies) or patients who are on specific medication (e.g. ACE inhibitors, statins, immunosuppressants).

In summary, this work demonstrates significant progress toward enabling scaled and digitized analyses of longitudinal unstructured and structured EHRs to identify variables (e.g. laboratory results) which are associated with relevant clinical phenotypes (e.g. COVID-19 diagnosis and outcomes). In doing so, we identified trends in lab test results which may be relevant to monitor in COVID-19 patients and warrant both clinical and mechanistic follow-up in more targeted and explicitly controlled prospective analyses.

Materials and methods

Study design, setting and patient population

Request a detailed protocol

This is a retrospective study of patients who underwent polymerase chain reaction (PCR) testing for suspected SARS-CoV-2 infection at the Mayo Clinic and hospitals affiliated to the Mayo health system. This research was conducted under IRB 20–003278, ‘Study of COVID-19 patient characteristics with augmented curation of Electronic Health Records (EHR) to inform strategic and operational decisions’. For further information regarding the Mayo Clinic Institutional Review Board (IRB) policy, and its institutional commitment, membership requirements, review of research, informed consent, recruitment, vulnerable population protection, biologics, and confidentiality policy, please refer to www.mayo.edu/research/institutional-review-board/overview.

Longitudinal lab testing tied to COVID-19 PCR diagnostic testing

Request a detailed protocol

We analyzed data from 74,586 patients who received PCR tests from the Mayo Clinic between February 15, 2020 to May 28, 2020. Among this population, 2232 patients had at least one positive SARS-CoV-2 PCR test result, and 72,354 patients had all negative PCR test results. In order to align the data for the analysis of aggregate longitudinal trends, we selected a reference date for each patient. For patients in the COVIDpos cohort, we used the date of the first positive PCR test result as the reference date (day = 0). For patients with all negative PCR tests, we used the date of the first PCR test result as the reference date (day = 0). We defined the study period for each patient to be 30 days before and after the PCR testing date. Patients with contradictory PCR test results were excluded for the purpose of this analysis; for example, a positive PCR test result and a negative PCR test result on the same day, or a positive PCR test result followed immediately by several negative PCR test results.

Over 4 million test results from 6298 different types of lab tests were recorded for the patients who received PCR tests in the 60-day window surrounding their PCR testing dates at the Mayo Clinic campuses in Minnesota, Arizona, and Florida. Among these lab tests, we restricted our analysis to 194 tests with at least 1000 observations total and at least 10 observations from the COVIDpos cohort among the patients with PCR testing on or before May 8, 2020. In addition, we considered different subsets of the COVIDpos cohort for the analysis of each of the 194 lab tests, due to differences in availability of testing results. For each lab test, we consider the results from patients with three or more observations during the study period.

In the end, there are 246 SARS-COV-2 positive and 13,666 SARS-CoV-2 negative patients that had three or more test results during the study period for at least one of the assays among the 194 lab tests considered. We take this set of 246 COVID-19 positive patients to be the COVIDpos cohort. In order to construct the COVIDneg cohort from the 13,666 COVID-19 negative patients, we apply propensity score matching, which is described in the next section.

Propensity score matching to select the final COVIDneg cohort

Request a detailed protocol

To construct a COVIDneg cohort similar in baseline clinical covariates to the COVIDpos cohort, we employ 1:10 propensity score matching (Austin, 2011). In particular, first we trained a regularized logistic regression model to predict the likelihood that each patient will have a positive or negative COVID-19 test result, using the following covariates: demographics (age, gender, race), anticoagulant/antiplatelet medication use (orders for alteplase, antithrombin III, apixaban, argatroban, aspirin, bivalirudin, clopidogrel, dabigatran, dalteparin, enoxaparin, eptifibatide, heparin, rivaroxaban, warfarin in the past year and in the past 30 days), pre-existing coagulopathies (medical history of thrombotic phenotypes including: deep vein thrombosis, pulmonary embolism, myocardial infarction, venous thromboembolism, thrombotic stroke, cerebral venous thrombosis, and disseminated intravascular coagulation from day −365 to day −31 relative to the PCR testing date), and hospitalization status (i.e. whether or not the patient was hospitalized within the past 30 days of PCR testing).

Using the predictions from the logistic regression model as propensity scores, we then matched each of the 246 patients in the COVIDpos cohort to 10 patients out of the 13,666 COVID-19 negative patients, using greedy nearest-neighbor matching without replacement (Austin, 2011; Austin, 2014). As a result, we ended up with a final COVIDneg cohort that included 2460 patients with similar baseline characteristics to the COVIDpos cohort. The characteristics of the two cohorts are summarized in Table 1.

Further, for the analyses conducted on individual lab tests, which include only a subset of patients from the COVIDpos cohort, we use the propensity scores to match each patient from the COVIDpos cohort to 10 patients from the COVIDneg cohort which have the most similar propensity scores and lab tests available. For example, for the fibrinogen lab test, in which we have data on 81 patients from the COVIDpos cohort, we select 810 patients from the COVIDneg cohort and the most similar propensity scores to be the control group. In this way, we ensure that all of the comparisons are done between subsets of the positive and negative cohorts with similar propensity scores, and therefore similar underlying characteristics.

Statistical significance assessments for lab test differences over prognostic time intervals for SARS-CoV-2 infection

Request a detailed protocol

We conduct a systematic statistical analysis to identify tests that show significant differentiation among the COVIDpos cohort during a set of predetermined prognostic time intervals for SARS-CoV-2 infection. In particular, we group the lab test measurements for each patient into the following nine time intervals relative to their date of PCR testing: pre-infection (days −30 to −11), pre-PCR (days −10 to −2), time of clinical presentation (days −1 to 0), and post-PCR phases 1 (days 1 to 3), 2 (days 4 to 6), 3 (days 7 to 9), 4 (days 10 to 12), 5 (days 13 to 15), and 6 (days 16 to 30).

For each lab test and for each of each of our nine pre-specified time intervals, we compared the mean lab test value among patients who underwent at least one such lab test in the COVIDpos cohort over that time interval to the mean lab test value in the COVIDneg (matched) cohort over that time window. We only considered (lab test, time interval) pairs in which there were at least three patients contributing to laboratory test results in both groups. Specifically, for each (lab test, time interval) pair, we conducted the following procedure:

  1. Compute (patient, time interval) averages: We compute the average lab test values for each patient in the COVIDpos and COVIDneg (matched) cohorts during the specified time interval.

  2. Statistical hypothesis testing: We conduct a Mann-Whitney U test in order to test the null hypothesis that the average lab test results for each of the (patient, time interval) pairs from the COVIDpos and COVIDneg (matched) cohorts come from the same distribution. In addition, we compute the Cohen’s D statistic as a measure of the effect size.

Once we have the statistics and p-values for each (test, time window) pair, in order to account for multiple hypotheses, we apply the Benjamini-Hochberg (BH) procedure with FDR controlled at 0.05. The results from the systematic comparisons which met our thresholds for effect size and statistical significance (Cohen’s D > 0.35, BH-adjusted Mann-Whitney p-value <0.05) are shown in Table 2.

Sensitivity analysis to assess the impact of perturbed clinical time windows

Request a detailed protocol

We perform a sensitivity analysis to assess whether or not the key findings from the systematic statistical assessment remain the same if we perturb the considered time intervals. In particular, we repeat the statistical analysis with the time intervals shifted forward or backward 1 day for all patients. For the forward shifted sensitivity analysis, the new time intervals under consideration are: pre-infection (days −30 to −10), pre-PCR (days −9 to −1), time of clinical presentation (days 0 to 1), and post-PCR phases 1 (days 2 to 4), 2 (days 5 to 7), 3 (days 8 to 10), 4 (days 11 to 13), 5 (days 14 to 16), and 6 (days 17 to 30). For the backward shifted sensitivity analysis, the new time intervals under consideration are: pre-infection (days −30 to −12), pre-PCR (days −11 to −3), time of clinical presentation (days −2 to −1), and post-PCR phases 1 (days 0 to 2), 2 (days 3 to 5), 3 (days 6 to 8), 4 (days 9 to 11), 5 (days 12 to 14), and 6 (days 15 to 30). For both the forward and backward sensitivity analyses, we apply the same thresholds of effect size and significance (Cohen’s D > 0.35, BH-adjusted Mann-Whitney p-value <0.05), and we compare the results to the original time intervals.

From this analysis, we observe consistent results (i.e. comparisons meeting same criteria of significance and effect) on (i) both perturbations in 83 out of 130 (64%) lab test trends identified in Table 2 and (ii) at least one perturbation in 114 of 130 (87%) lab test trends. In Table 3, we report the specific results of the time shifted windows for five coagulation-related lab tests (fibrinogen, platelets, prothrombin time, activated partial thromboplastin time, and D-dimer).

Augmented curation of anticoagulant administration and the coagulopathy outcomes from the unstructured clinical notes and their triangulation to structured EHR databases

Request a detailed protocol

A state-of-the-art BERT-based neural network (Devlin et al., 2018) was previously developed to classify sentiment regarding a diagnosis in the EHR (Wagner et al., 2020). Sentences containing phenotypes were classified into the following categories: Yes (confirmed diagnosis), No (ruled out diagnosis), Maybe (possibility of disease), and Other (alternate context, e.g. family history of disease). The neural network used to perform this classification was trained using nearly 250 different phenotypes and 18,500 sentences and achieves 93.6% overall accuracy and over 95% precision and recall for Yes/No sentiment classification (Wagner et al., 2020). Here, this model was used to classify the sentiment around coagulopathies in the unstructured text of the 246 COVIDpos and 13,666 COVIDneg patients’ clinical notes, structuring this information so that it could be compiled with longitudinal lab measurement and medication information.

In particular, we used the BERT model to identify the seven coagulopathy phenotypes mentioned in clinical notes in the Mayo Clinic EHR database, including: deep vein thrombosis, pulmonary embolism, myocardial infarction, venous thromboembolism, thrombotic stroke, cerebral venous thrombosis, and disseminated intravascular coagulation. We validated the performance of this model for these phenotypes on a set of 1000 randomly selected sentences from the clinical notes of the patients in the study population. In Table 6, we report the out-of-sample accuracy metrics for the BERT model on this set of sentences, using manually curated labels provided by one of the study’s authors (CP) to be the ground truth. We demonstrate that the model performs well in the task of identifying thrombotic phenotypes in clinical notes, with an overall accuracy of 94.7%, recall of 97.8%, and precision of 92.8%.

References

  1. 1
  2. 2
  3. 3
    (COVID-19) - Symptoms and Causes - Mayo Clinic. (2020)
    1. Coronavirus disease
    (2019)
    Mayo clinic.
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
    Symptoms of coronavirus
    1. Website
    (2020)
    Accessed May 13, 2020.
  14. 14
  15. 15
  16. 16

Decision letter

  1. Jos WM van der Meer
    Senior Editor; Radboud University Medical Center, Netherlands
  2. Frank L van de Veerdonk
    Reviewing Editor; Radboud University Medical Center, Netherlands
  3. Coen Maas
    Reviewer; University Medical Center Utrecht, Netherlands
  4. Jinbo Chen
    Reviewer; University of Pennsylvania, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The paper analyzes a large EHR-based dataset of coagulation tests in COVID-19 patients to obtain an understanding of the kinetics of COVID-19 associated coagulopathy. By using machine learning to extract patterns data are provided which support that the majority of thrombotic events in COVID-19 patients are not the result of a DIC-like consumptive coagulopathy, and that this only occurs in a small subset.

Decision letter after peer review:

Thank you for submitting your article "Longitudinal laboratory testing tied to PCR diagnostics in COVID-19 patients reveals temporal evolution of coagulopathy" for consideration by eLife. Your article has been reviewed by four peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Jos van der Meer as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Coen Maas (Reviewer #1); Jinbo Chen (Reviewer #4).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data.

Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Summary:

The authors analyze a large EHR-based dataset of coagulation tests in COVID-19 patients to obtain an understanding of the kinetics of COVID-19 associated coagulopathy. It is a clear approach and the data are important for the clinics. They used machine learning to extract patterns. Overall the conclusions were drawn in many directions. Nonetheless, the paper asks a potentially important question and is adequately written and presented. The data support that the majority of thrombotic events in COVID-19 patients are not the result of a DIC-like consumptive coagulopathy, this only occurs in a small subset. This should also be the focus of the Discussion.

Essential revisions:

Patient characteristics:

– Please clinically define "pre-existing coagulopathies".

– It is a unclear what the comparator cohort adds: In part it is not clear who these patients are. Perhaps the controls should be age/sex-matched patients with other causes of hypoxemic respiratory failure/pneumonia? Or different COVID-19 disease severity subgroups? Otherwise what are we inferring from comparing these two groups?

– Ascertained rates of venous thromboembolism are much lower than what others have observed (e.g., Helms ICM 2020, Middeldorp pre-prints, etc.). Could this relate to ascertainment bias in this cohort? Perhaps a push not to perform diagnostic imaging studies lead to lower objective identification of venous thromboembolism than has been generally described in COVID-19?

– Patients are identified as COVID-19 positive or negative based on PCR testing. Actually, using the PCR test a positive/negative infection by SARS-CoV-2 is assessed; COVID-19 is the disease that can follow upon infection. This distinction should be made clear, as not all patients tested positive for SARS-CoV-2 infection necessarily develop COVID-19. What was the guideline followed for PCR-testing: displaying COVID-19 symptoms/ contact with infected persons/ other? This should also be explained and included.

– Introduction: "… straddling the date of the PCR test.…" Do the authors refer to the first PCR test here, as it is mentioned that multiple PCR tests may have been performed? Please include a definition.

Analysis:

– The analysis was restricted to patients who had serial (>/= 3) tests done. This could have led to a survival bias which should be acknowledged.

– This restriction also led to the exclusion of the vast majority of the cohort: 1,192 -> 181 COVID-19pos patients. Why was there such a strong emphasis on longitudinal markers?

– The potential significance of these longitudinal trends are assessed with these tests as far as I can understand, which both only looks to see whether mean/median change was different between the groups, but moreover does not allow for adjustment. Why not use multilevel regression mixture models, longitudinal regression, etc?

– One of the three covariates used in the regularized logistic regression model to predict the likelihood of a positive infection (subsection “Propensity score matching to select the final COVIDneg cohort”) is anticoagulant/antiplatelet medication use. Is this covariate positively or negatively correlated to infection and what is the rationale for this? Table 1 comprises more details; however, these add to further confusion. The Table 1 legend states "anticoagulant/antiplatelet use within 30 days/1 year of PCR testing date" i.e. after the PCR test, while the table itself mentions "medication use in the preceding 30 days/1yr", so before the PCR test. Please clarify and adapt.

Figures:

– Figure 2A: Why are no data points provided for the -30 to 0 days of the COVID negative patients while these data are shown in Figure 2B and E?

– The same question as above but then for the APTT and D-dimer data in Figure 2 vs. Figure 3 and magnesium in Figure 1 vs. Figure 3. I guess the cohorts are different between the two figures, maybe this should be stated in the figure legends.

– Figure 3: we agree with the authors that the fibrinogen decline and platelet increase in COVID positive patients is re-emphasized in this manner. Also the increase in magnesium and decrease in alkaline phosphatase seem to stand out. Could the authors comment on this?

– Could the authors comment on the number of thrombotic events that were radiographically-confirmed?

– Figure 5: for the individual patients could the authors comment on the heparin therapy with regard to the source of heparin (LMWH vs. UFH) and dosing (prophylactic vs. therapeutic)?

– Figure 6 is very hard to understand; please find another way to graphically display the findings in a clear manner.

Specific statistical comments:

1) Cohort identification and description. (a) For the 1.3 million lab tests on 194 assays over the 60-day window: provide the mean/range of number of tests separately for positive and negative patients, and also provide the mean/range of the number of tests for pre- vs. post-index "0" date; (b) For positive patients, the proportion of those whose PCR positive test is the first test in the 60-day window; (c) If at all possible, COVID-19 related info at the first PCR test for all patients; (4) If possible, reasons for hospitalizations at the day=0;

2) It is not clear why it is important to assess lab tests in relation to diagnosis. For studying association between lab tests with diagnosis, ideally, the positive cases should use the "time of infection" as time zero which of course is intractable. But before variation in time from infection to diagnosis by PCR tests, the relevance of test results for diagnosis is unclear. Further, because of the propensity score matching, the cohort may not be suitable for assess the diagnosis: if some matching variables are associated with any of the test results, matching will artificially deflate the association. We therefore suggest the focus of this paper on prognosis only. The Abstract indeed focused on the prognosis, but diagnosis was mentioned multiple times in the manuscript.

3) BERT method: The accuracy of BERT was established in an unpublished manuscript developed by the same research group. What is the proportion of patients who were classified as "Maybe", "Yes" AND "No"? It is helpful if chart review is performed on 100, say, patients, to validate this algorithm in the study context;

4) Tables 3 and 4 indicate that the endpoints were significantly enriched for positive cases who had longitudinal data. But it is possible that patients received more tests because of indications of more negative outcomes. As a result, patients with longitudinal lab tests are representative of all positive patients as indicated by the enrichment. This is an important weakness of the study as this may be a source of bias, making the study results not generalizable. It is important to provide additional information on the comparing characteristics of patients with and without longitudinal tests. It may be worthwhile to provide information on reasons of repeating tests for negative patients.

https://doi.org/10.7554/eLife.59209.sa1

Author response

Summary:

The authors analyze a large EHR-based dataset of coagulation tests in COVID-19 patients to obtain an understanding of the kinetics of COVID-19 associated coagulopathy. It is a clear approach and the data are important for the clinics. They used machine learning to extract patterns. Overall the conclusions were drawn in many directions. Nonetheless, the paper asks a potentially important question and is adequately written and presented. The data support that the majority of thrombotic events in COVID-19 patients are not the result of a DIC-like consumptive coagulopathy, this only occurs in a small subset. This should also be the focus of the Discussion.

We thank the editors for their advice to focus the key conclusion of the paper. Accordingly, we have revised the manuscript’s Discussion section to reflect this suggestion. We find that the updated data for this manuscript also support this conclusion that DIC-like consumptive coagulopathy is relatively rare among COVID-19 patients, as shown in updated Table 3.

Essential revisions:

Patient characteristics:

– Please clinically define "pre-existing coagulopathies".

We consider the following phenotypes to be coagulopathies: deep vein thrombosis, pulmonary embolism, myocardial infarction, venous thromboembolism, thrombotic stroke, cerebral venous thrombosis, and disseminated intravascular coagulation. We have added a definition of pre-existing coagulopathies to the Introduction, which includes all patients who have at least one of the phenotypes from days -365 to -31 relative to the PCR testing date. In Table 1 of the revised manuscript, we have now added in the medical history of thrombotic events for the patients in the COVID-19 positive, negative, and negative (matched) cohorts.

– It is a unclear what the comparator cohort adds: In part it is not clear who these patients are. Perhaps the controls should be age/sex-matched patients with other causes of hypoxemic respiratory failure/pneumonia? Or different COVID-19 disease severity subgroups? Otherwise what are we inferring from comparing these two groups?

The comparator cohort was selected to match the COVIDpos cohort on a variety of clinical characteristics which may be confounding variables for thrombotic events and related lab tests that we observe during the study period. The clinical covariates that we controlled for initially included demographics (age, gender, race), hospitalization status, and anticoagulant/antiplatelet use, which can be viewed as a proxy variable for medical history of thrombotic events. In this revision, we have explicitly added in medical history of thrombotic events as clinical covariates to balance on, as shown in the updated Table 1 of the revised manuscript. We have added a few statements describing the updated COVIDneg (matched) cohort to the last paragraph of the Introduction.

– Ascertained rates of venous thromboembolism are much lower than what others have observed (e.g., Helms ICM 2020, Middeldorp pre-prints, etc.). Could this relate to ascertainment bias in this cohort? Perhaps a push not to perform diagnostic imaging studies lead to lower objective identification of venous thromboembolism than has been generally described in COVID-19?

Our study presents, for the first time, fine grained temporal resolution into lab results change in all COVIDpos patients. In particular, our study is not restricted to severe MODS or ARDS patients, as has been the focus of previous studies such as Helms ICM 2020, Middeldorp preprints, or https://pubmed.ncbi.nlm.nih.gov/32353745/ etc. We expect that this is the reason for the differences in venous thromboembolism rate reported across the different studies.

– Patients are identified as COVID-19 positive or negative based on PCR testing. Actually, using the PCR test a positive/negative infection by SARS-CoV-2 is assessed; COVID-19 is the disease that can follow upon infection. This distinction should be made clear, as not all patients tested positive for SARS-CoV-2 infection necessarily develop COVID-19. What was the guideline followed for PCR-testing: displaying COVID-19 symptoms/ contact with infected persons/ other? This should also be explained and included.

As recommended by the reviewers, the distinction between SARS-CoV-2 infection and the COVID-19 disease onset and progression has been clarified in the revised paper. The guidelines followed for PCR-testing included all of the following – displaying COVID-19 symptoms, contact with infected persons, and other factors such as potential work-place/community exposure as well as underlying conditions predisposing to SARS-CoV-2 infection. This has been noted in the second paragraph of the Introduction section of the manuscript.

– Introduction: "… straddling the date of the PCR test.…" Do the authors refer to the first PCR test here, as it is mentioned that multiple PCR tests may have been performed? Please include a definition.

We have added a statement to the Introduction, clarifying this point. For the COVIDpos cohort, we center the two month observation period around the date of the first positive PCR test for SARS-CoV-2, and for the COVIDneg cohort, we center the two month observation period around the date of the first PCR test for SARS-CoV-2.

Analysis:

– The analysis was restricted to patients who had serial (>/= 3) tests done. This could have led to a survival bias which should be acknowledged.

We thank the reviewers for raising this point. We have added in a statement on potential survival bias to the Discussion section. In particular, we now provide a summary of the clinical characteristics of the study population vs. all patients with PCR tests during the same time period in Table 6. We have also included an additional table characterizing the key lab tests and their availability for the study population vs. all patients (see Tables 7-8).

– This restriction also led to the exclusion of the vast majority of the cohort: 1,192 -> 181 COVID-19pos patients. Why was there such a strong emphasis on longitudinal markers?

The need for longitudinal data on the testing results, while constraining the study population size greatly, enables us to provide a fine-grained temporal resolution of COVID-associated coagulopathy for the first time. In addition, we consider this study population to be of high clinical interest because these patients are highly enriched for severe thrombotic events during the study period (see Table 4). We have added this in as a point in the seventh paragraph of the Discussion section.

– The potential significance of these longitudinal trends are assessed with these tests as far as I can understand, which both only looks to see whether mean/median change was different between the groups, but moreover does not allow for adjustment. Why not use multilevel regression mixture models, longitudinal regression, etc?

Thank you for the suggestion. We considered logistic regression for this analysis, to predict the presence/absence of thrombotic events using the longitudinal lab testing information available for each patient prior to the PCR testing date Day 0. However, due to large heterogeneity in lab tests conducted for the study population, and the relative sparsity of lab testing data available for the COVIDpos population prior to Day 0, we were unable to train an accurate predictive model using the available data. However, we expect that as additional data becomes available this is a promising area for future research, and we have added a note on this in the Discussion section.

– One of the three covariates used in the regularized logistic regression model to predict the likelihood of a positive infection (subsection “Propensity score matching to select the final COVIDneg cohort”) is anticoagulant/antiplatelet medication use. Is this covariate positively or negatively correlated to infection and what is the rationale for this? Table 1 comprises more details; however, these add to further confusion. The Table 1 legend states "anticoagulant/antiplatelet use within 30 days/1 year of PCR testing date" i.e. after the PCR test, while the table itself mentions "medication use in the preceding 30 days/1yr", so before the PCR test. Please clarify and adapt.

We observe that anticoagulant/antiplatelet use is negatively correlated with SARS-CoV-2 infection (see Table 1). We hypothesize that the reason for this is that patients who are on anticoagulants/antiplatelets generally have more underlying conditions and thus take more precautionary measures to avoid SARS-CoV-2 infection. In addition, Table 1 shows anticoagulant/antiplatelet use in the 30 days/1 year prior to the PCR testing date. We have revised the name of this covariate in Table 1 for clarification.

Figures:

– Figure 2A: Why are no data points provided for the -30 to 0 days of the COVID negative patients while these data are shown in Figure 2B and E?

In the revised manuscript, Figure 2A has been renamed to Figure 3A. This plot, which shows the temporal trends of the Fibrinogen lab tests, has no data points between days -30 to 0 because none of the patients in the COVIDpos cohort had Fibrinogen lab test results reported during this time range. For all of the temporal trend plots, we only show data points for which there are at least 3 patients in the COVIDpos and COVIDneg cohorts. We have clarified this in the legends for the updated Figures 1, 2, and 3, and 4.

– The same question as above but then for the APTT and D-dimer data in Figure 2 vs. Figure 3 and magnesium in Figure 1 vs. Figure 3. I guess the cohorts are different between the two figures, maybe this should be stated in the figure legends.

The cohorts are different because the previous Figure 3 (now Figure 4) is restricted to patients with 3 or more lab tests during the study period, while the previous Figures 1 and 2 (now Figures 1, 2, 3) include patients with 1 or more lab tests during the study period. We have added this statement to the figure legends for clarification.

– Figure 3: we agree with the authors that the fibrinogen decline and platelet increase in COVID positive patients is re-emphasized in this manner. Also the increase in magnesium and decrease in alkaline phosphatase seem to stand out. Could the authors comment on this?

We agree that the increases in magnesium and alkaline phosphatase do seem significant. We are planning to explore this trend in future analyses.

– Could the authors comment on the number of thrombotic events that were radiographically-confirmed?

We do not have the radiology reports available in this dataset, so unfortunately we cannot comment on the number of radiographically confirmed events.

– Figure 5: for the individual patients could the authors comment on the heparin therapy with regard to the source of heparin (LMWH vs. UFH) and dosing (prophylactic vs. therapeutic)?

We have added annotations to Figure 5 indicating which medications were prescribed for prophylaxis. In addition, we have updated the Results section to include a note on these results.

– Figure 6 is very hard to understand; please find another way to graphically display the findings in a clear manner.

In order to keep the paper focused on clinical data, we have now excluded the previous Figure 6, which was based on single cell data.

Specific statistical comments:

1) Cohort identification and description. (a) For the 1.3 million lab tests on 194 assays over the 60-day window: provide the mean/range of number of tests separately for positive and negative patients, and also provide the mean/range of the number of tests for pre- vs. post-index "0" date; (b) For positive patients, the proportion of those whose PCR positive test is the first test in the 60-day window; (c) If at all possible, COVID-19 related info at the first PCR test for all patients; (4) If possible, reasons for hospitalizations at the day=0;

We include Tables 7 and 8 which contain the relevant numbers for (1) (a)-(b); this should hopefully provide some further context. For PCR-positive and PCR-negative patient cohorts, we include the overall number of tests and number of patients with tests; we do this for all the tests in aggregate and 5 coagulation-related tests in particular. Table 7 has these numbers for the PCR-tested patient population; Table 8 has them for the study population of PCR-tested patients with longitudinal test data. We included general demographic characteristics as well in Table 6. We were not able to address (c) further or (d) at this time unfortunately.

2) It is not clear why it is important to assess lab tests in relation to diagnosis. For studying association between lab tests with diagnosis, ideally, the positive cases should use the "time of infection" as time zero which of course is intractable. But before variation in time from infection to diagnosis by PCR tests, the relevance of test results for diagnosis is unclear. Further, because of the propensity score matching, the cohort may not be suitable for assess the diagnosis: if some matching variables are associated with any of the test results, matching will artificially deflate the association. We therefore suggest the focus of this paper on prognosis only. The Abstract indeed focused on the prognosis, but diagnosis was mentioned multiple times in the manuscript.

We have clarified that the focus of this paper on prognosis rather than diagnosis. Furthermore, only explicit statements “at the time of diagnosis”, “after diagnosis”, or “following diagnosis” are retained in the Discussion and Results section. This should address the key concern from a statistical analysis standpoint above.

3) BERT method: The accuracy of BERT was established in an unpublished manuscript developed by the same research group. What is the proportion of patients who were classified as "Maybe", "Yes" AND "No"? It is helpful if chart review is performed on 100, say, patients, to validate this algorithm in the study context;

We have added in validation statistics for the BERT model on a randomly selected set of 1,000 sentences with thrombotic event phenotypes (see Table 5). Overall, we observe that the BERT model has an out-of-sample accuracy of 94.7%, recall of 97.8%, and precision of 92.8%. We have added a paragraph to the Materials and methods section about this validation procedure and the results.

4) Tables 3 and 4 indicate that the endpoints were significantly enriched for positive cases who had longitudinal data. But it is possible that patients received more tests because of indications of more negative outcomes. As a result, patients with longitudinal lab tests are representative of all positive patients as indicated by the enrichment. This is an important weakness of the study as this may be a source of bias, making the study results not generalizable. It is important to provide additional information on the comparing characteristics of patients with and without longitudinal tests. It may be worthwhile to provide information on reasons of repeating tests for negative patients.

We have added in a table showing the demographics of the patients who received PCR testing and our study population of patients who received PCR testing and took at least one lab test on 3+ separate days (Table 6A and B respectively). In addition, we agree with the reviewer that the study cohort of patients with longitudinal lab tests is enriched for patients with more severe disease outcomes. However, in this revision we have added in medical history clinical covariates to control for in the propensity score matching step of this analysis, so we believe that the COVIDpos and COVIDneg (matched) to be similar in their propensity for more severe disease outcomes.

https://doi.org/10.7554/eLife.59209.sa2

Article and author information

Author details

  1. Colin Pawlowski

    nference, inc, Cambridge, United States
    Contribution
    Formal analysis, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    This author is an employee of nference with financial interests in the company
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2781-7507
  2. Tyler Wagner

    nference, inc, Cambridge, United States
    Contribution
    Formal analysis, Supervision, Validation, Investigation, Methodology, Writing - original draft, Project administration, Writing - review and editing
    Competing interests
    This author is an employee of nference with financial interests in the company
  3. Arjun Puranik

    nference, inc, Cambridge, United States
    Contribution
    Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - review and editing
    Competing interests
    This author is an employee of nference with financial interests in the company
  4. Karthik Murugadoss

    nference, inc, Cambridge, United States
    Contribution
    Software, Formal analysis, Validation, Investigation, Methodology, Writing - review and editing
    Competing interests
    This author is an employee of nference with financial interests in the company
  5. Liam Loscalzo

    nference, inc, Cambridge, United States
    Contribution
    Formal analysis, Supervision, Investigation, Methodology, Project administration
    Competing interests
    This author is an employee of nference with financial interests in the company
  6. AJ Venkatakrishnan

    nference, inc, Cambridge, United States
    Contribution
    Formal analysis, Validation, Investigation, Methodology, Writing - review and editing
    Competing interests
    This author is an employee of nference with financial interests in the company
  7. Rajiv K Pruthi

    Mayo Clinic, Rochester, United States
    Contribution
    Supervision, Validation, Investigation, Methodology, Writing - original draft, Project administration, Writing - review and editing
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. Mayo Clinic is an investor in nference. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies.
  8. Damon E Houghton

    Mayo Clinic, Rochester, United States
    Contribution
    Formal analysis, Supervision, Validation, Investigation, Methodology, Project administration, Writing - review and editing
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. Mayo Clinic is an investor in nference. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies.
  9. John C O'Horo

    Mayo Clinic, Rochester, United States
    Contribution
    Supervision, Investigation, Project administration, Writing - review and editing
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. Mayo Clinic is an investor in nference. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies.
  10. William G Morice II

    1. Mayo Clinic, Rochester, United States
    2. Mayo Clinic Laboratories, Rochester, United States
    Contribution
    Supervision, Investigation, Methodology, Project administration
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies. The author is also involved in the Mayo Clinic Laboratories.
  11. Amy W Williams

    Mayo Clinic, Rochester, United States
    Contribution
    Supervision, Investigation, Project administration, Writing - review and editing
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. Mayo Clinic is an investor in nference. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies.
  12. Gregory J Gores

    Mayo Clinic, Rochester, United States
    Contribution
    Supervision, Investigation, Project administration, Writing - review and editing
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. Mayo Clinic is an investor in nference. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies.
  13. John Halamka

    1. Mayo Clinic, Rochester, United States
    2. Mayo Clinic Platform, Rochester, United States
    Contribution
    Supervision, Project administration
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. Mayo Clinic is an investor in nference. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies.
  14. Andrew D Badley

    Mayo Clinic, Rochester, United States
    Contribution
    Supervision, Validation, Project administration, Writing - review and editing
    Competing interests
    One or more of the investigators associated with this project and Mayo Clinic have a Financial Conflict of Interest in technology used in the research and that the investigator(s) and Mayo Clinic may stand to gain financially from the successful outcome of the research. Mayo Clinic is an investor in nference. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and is being conducted in compliance with Mayo Clinic Conflict of Interest policies.
  15. Elliot S Barnathan

    Janssen pharmaceutical companies of Johnson & Johnson (J&J), Spring House, United States
    Contribution
    Supervision, Validation, Investigation, Methodology, Writing - review and editing
    Competing interests
    This author is an employee of the Janssen pharmaceutical companies of J&J with financial interests in the company
  16. Hideo Makimura

    Janssen pharmaceutical companies of Johnson & Johnson (J&J), Spring House, United States
    Contribution
    Supervision, Validation, Investigation, Methodology, Writing - review and editing
    Competing interests
    This author is an employee of the Janssen pharmaceutical companies of J&J with financial interests in the company
  17. Najat Khan

    Janssen pharmaceutical companies of Johnson & Johnson (J&J), Spring House, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Validation, Investigation, Methodology, Writing - review and editing
    Competing interests
    This author is an employee of the Janssen pharmaceutical companies of J&J with financial interests in the company
  18. Venky Soundararajan

    nference, inc, Cambridge, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    venky@nference.net
    Competing interests
    The author is an employee of nference and has financial interests in the company. Outside the submitted work, Venky Soundararajan is listed as inventor of the following patent: "Systems, methods, and computer readable media for visualization of semantic information and inference of temporal signals indicating salient associations between life science entities" (US20180082197A1).
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7434-9211

Funding

No external funding was received for this work.

Acknowledgements

The authors thank Mathai Mammen, James List, JoAnne Foody, Patrick Lenehan, Murali Aravamudan, Rakesh Barve, Sankar Ardhanari, and Vishy Thiagarajan, for their helpful feedback.

Ethics

Human subjects: This research was conducted under IRB 20-003278, "Study of COVID-19 patient characteristics with augmented curation of Electronic Health Records (EHR) to inform strategic and operational decisions". All analysis of EHRs was performed in the privacy-preserving environment secured and controlled by the Mayo Clinic. nference, the Mayo Clinic, and the Janssen pharmaceutical companies of Johnson & Johnson (J&J) subscribe to the basic ethical principles underlying the conduct of research involving human subjects as set forth in the Belmont Report and strictly ensure compliance with the Common Rule in the Code of Federal Regulations (45 CFR 46) on the Protection of Human Subjects.

Senior Editor

  1. Jos WM van der Meer, Radboud University Medical Center, Netherlands

Reviewing Editor

  1. Frank L van de Veerdonk, Radboud University Medical Center, Netherlands

Reviewers

  1. Coen Maas, University Medical Center Utrecht, Netherlands
  2. Jinbo Chen, University of Pennsylvania, United States

Publication history

  1. Received: May 22, 2020
  2. Accepted: August 14, 2020
  3. Accepted Manuscript published: August 17, 2020 (version 1)
  4. Version of Record published: September 4, 2020 (version 2)

Copyright

© 2020, Pawlowski et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 480
    Page views
  • 92
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

  1. Edited by Jos WM van der Meer et al.
    Collection

    eLife has published the following articles on SARS-CoV-2 and COVID-19.

    1. Genetics and Genomics
    2. Medicine
    Kristoffer B Hansen et al.
    Research Article

    Acid-base conditions modify artery tone and tissue perfusion but the involved vascular sensing mechanisms and disease consequences remain unclear. We experimentally investigated transgenic mice and performed genetic studies in a UK-based human cohort. We show that endothelial cells express the putative HCO3-sensor receptor-type tyrosine-protein phosphatase RPTPg, which enhances endothelial intracellular Ca2+-responses in resistance arteries and facilitates endothelium-dependent vasorelaxation only when CO2/HCO3 is present. Consistent with waning RPTPg-dependent vasorelaxation at low [HCO3], RPTPg limits increases in cerebral perfusion during neuronal activity and augments decreases in cerebral perfusion during hyperventilation. RPTPg does not influence resting blood pressure but amplifies hyperventilation-induced blood pressure elevations. Loss-of-function variants in PTPRG, encoding RPTPg, are associated with increased risk of cerebral infarction, heart attack, and reduced cardiac ejection fraction. We conclude that PTPRG is an ischemia susceptibility locus; and RPTPg-dependent sensing of HCO3 adjusts endothelium-mediated vasorelaxation, microvascular perfusion, and blood pressure during acid-base disturbances and altered tissue metabolism.