1 INTRODUCTION

Disseminated intravascular coagulation (DIC) is characterized by systemic coagulation and platelet activation that results in global microthrombus formation and organ dysfunction. Further, concomitant consumption of platelets and coagulation factors can lead to life-threatening bleeding.1, 2 The diagnosis of DIC encompasses findings of coagulopathy or disturbed fibrinolysis in the presence of an underlying cause of DIC (eg, sepsis or malignancy);3, 4 however, the clinical diagnosis of DIC is difficult, and therefore laboratory tests are vital for the diagnosis.5 As no single laboratory test can accurately confirm or eliminate the diagnosis of DIC, international guidelines recommend the use of DIC scoring systems to support the diagnosis of DIC.6, 7 The existing scoring systems are based on laboratory tests available in most hospital laboratories. One of the most widely employed scoring systems was developed by the ISTH. The score is based on the patient’s platelet count, international normalized ratio (INR), fibrinogen, and fibrin D-dimer. A score of ≥5 indicates overt DIC.5 Other scoring systems, such as the Japanese Association for Acute Medicine (JAAM) DIC score, include antithrombin but may exclude fibrinogen, and the JAAM DIC score also uses change in platelet count as a supplement to absolute platelet count.8 In the present study, we focused on the ISTH DIC score, as this scoring system is one of the most used in the clinical setting in the Western part of the world and because it can be calculated solely on the basis of laboratory database records, which represented our main data source.

DIC scoring systems are reported to have a sensitivity of ≈90% and a specificity of ≈97%.9, 10 However, the positive predictive value (PPV) of different DIC scoring systems will depend on the prevalence of DIC in the source population. Furthermore, as the score is laboratory based, it is sensitive to the presence of factors that can influence the laboratory parameters comprising the DIC score, for example, liver disease, which can cause prolonged prothrombin time and in advanced stages, thrombocytopenia or hematological disease, which can affect platelet count either directly or through antineoplastic treatment. Thus, the PPV of the DIC score must be expected to vary depending on the underlying cause of DIC, the clinical setting (internal medicine ward/intensive care unit [ICU]), the severity of coagulopathy (DIC score 5 vs 8), and the presence of coexisting conditions and medications that influence routine coagulation analyses (eg, liver dysfunction or vitamin K antagonist treatment).11 Furthermore, the ISTH DIC score was introduced 20 years ago this year, and the demography and clinical characteristics of patients with DIC may have changed during this time span. Estimating the PPV for the ISTH DIC score in a hospital-based setting is important for two reasons: First, it will assist clinicians in the evaluation of the patient with suspected DIC, ensuring timely diagnosis and interventions; second, valid identification of patients with DIC from laboratory information system databases will enable research on DIC epidemiology. The present study therefore examined the PPV of the ISTH DIC score for overt DIC in an adult Danish in-hospital population.

2 MATERIALS AND METHODS 2.1 Setting and design

This validation study was conducted in the Central Denmark Region from January 1, 2015, through December 31, 2018, within a source population of 1.2 million residents. Denmark has a tax-supported health care system that guarantees unfettered access to medical care for all residents.12

2.2 Study population

Patients were eligible for inclusion if (i) laboratory results for all ISTH DIC score components (platelet count, INR, fibrinogen, fibrin D-dimer) were available from the same date, (ii) they had a first-time ISTH DIC score ≥5, and (iii) the main indication for ordering the laboratory tests was a clinical suspicion of DIC and no other causes (eg, surgical bleeding).

We searched the hospital laboratory information system in the Central Denmark Region to sample adult patients (≥18 years old) with potential DIC diagnoses using a predefined algorithm. We identified patients from Aarhus University Hospital and two regional hospitals (Regional Hospital West Jutland and Regional Hospital Central Jutland). All biomarkers used at the hospital departments are tracked in the electronic hospital laboratory information system using the International System of Nomenclature, Properties, and Units.13 Given the homogeneity of the Danish health care system, these hospitals are considered representative for similar-sized hospitals in other Danish regions.14

First, we identified all patients with laboratory records of platelet count, INR, fibrin D-dimer, and fibrinogen levels during the study period (Figure 1 and Table S1). Then, we restricted to patients who had all components of the ISTH DIC scoring system measured on the same date. After this, we calculated all their ISTH DIC scores during the study period using the algorithms provided in Table 1. For patients with more measurements (eg, platelet count measures) on the same day, we used the most abnormal values, that is, the lowest value for measurements of platelet counts and fibrinogen and the highest value for fibrin D-dimer and INR. Patients were included only once, according to their first record of a given DIC score ≥5. Records from general practitioners, emergency room departments, and outpatients (likely not requested on DIC indication) were excluded from the analyses, but they were included if they were later admitted to a hospital department. We subsequently divided the study population into three groups with ISTH DIC scores of 5, 6, or ≥7.

Flowchart showing selection of patients from the laboratory database and the random samples from each DIC score category. DIC, disseminated intravascular coagulation; ER, emergency room; GPs, general practitioners; INR, international normalized ratio

TABLE 1. ISTH DIC score algorithms used in the study ISTH-DIC score Value Score

International normalized ratio

Reference range <1.2

<1.3 0 1.3-1.6 1 >1.6 2

Fibrin D-dimer (mg/L)

Reference range: age-specifica

<0.8 0 ≥0.8-8.0 2 >8.0 3

Platelet count (109/L)

Reference range 145–400

>100 0 50–100 1 <50 2

Fibrinogen (µmol/L)

Reference range: 5.5-12

≥2.94 (equivalent to ≥1 g/L) 0 <2.94 (equivalent to <1 g/L) 1 Abbreviation: DIC, disseminated intravascular coagulation. a 18-55 years:<0.50, 55-65 years:<0.60, 65-75 years:<0.70, 75-85 years:<0.80, 85-95 years: <0.90.

Some patients may have had testing performed on other indications than suspicion of DIC. To estimate the proportion of patients with DIC as an indication for testing, in a pilot project, we randomly selected 10 patients from each ISTH DIC score group and reviewed their medical records to determine the indication for DIC score testing. In patients with a DIC score of 5, a relevant indication was present in only 50% of patients. Based on these preliminary results, we decided to include a sample size of up to 200 patients in each ISTH DIC score group to obtain 75 patients with a relevant indication for DIC testing in each group. The sample size of 75 has shown to yield relatively precise effect estimates in prior validation studies.15, 16

2.3 Adjudication of the DIC diagnosis

A true gold standard for the clinical diagnosis of DIC does not exist. Therefore, the patients were classified as having overt (symptomatic) DIC, subclinical DIC, or no DIC through expert consensus, based on available information in the electronic medical record during a window of ±2 days within the time of blood sampling for DIC score. The diagnosis of overt DIC was based on the presence or strong suspicion of bleeding and/or microthrombus formation (acute organ dysfunction, blue extremities, necrosis/gangrene) in combination with relevant laboratory results not included in the ISTH DIC score (eg, activated partial thromboplastin time [aPTT], antithrombin, increment or decrement in platelet count), microbiology results, imaging results, information on medication, and the absence of other conditions that could clearly explain abnormal laboratory results (eg, liver cirrhosis in a patient with moderate thrombocytopenia and and prolonged INR). The diagnosis of subclinical DIC was based on clear biochemical signs of DIC and, as for overt DIC, an absence of other conditions that could clearly explain abnormal laboratory results, but no suspicion of bleeding or microthrombus formation. To ensure consensus on data collection and assignment of a clinical diagnosis of DIC, five medical records first were reviewed and discussed by all authors. Subsequently, one of three authors (JBL, MAA, or KA) reviewed each medical record and judged whether a definite diagnosis of overt DIC should be made. In case of doubt, cases were discussed among the three authors to reach consensus, and if the diagnosis remained uncertain, a secondary review was performed together with an expert in thrombosis and hemostasis (AMH) to reach consensus.

2.4 Medical record data collection

Data were systematically extracted from the medical records using the form provided in Table S2 and entered into REDCap (Research Electronic Data Capture; Vanderbilt University, Nashville, TN, USA) hosted by Central Denmark Region.17

Data items included (i) baseline characteristics (age and sex); (ii) indication for DIC testing; (iii) presence or suspicion of relevant DIC etiology; (iv) presence or suspicion of possible differential diagnosis to DIC; (v) all-cause mortality; (vi) presence or suspicion of venous or arterial thrombotic disease; (vii) presence or suspicion of thrombotic microangiopathy, bleeding, and circulatory instability; (viii) Simplified Acute Physiology Score at ICU admission, if available, and Sequential Organ Failure Assessment (SOFA) score; (ix) laboratory records; (x) microbiology results; (11) relevant treatments (Table S2). A time window of ±2 days was set for the classical clinical manifestations of DIC thrombotic microangiopathy and bleeding to allow for potential reporting delay in the medical record. Bleeding was defined and categorized according to the modified World Health Organization (WHO) bleeding score18 (Table S2). The final decision on presence/absence of overt DIC also took into account presence/absence of circulatory instability, presence of other conditions, which could influence the laboratory parameters included in the DIC score, and other laboratory results (eg, aPTT and antithrombin).

2.5 Statistical analysis

Only patients with a relevant indication for DIC testing were included in the analysis. For all patients in total and for each subscore (ie, ISTH DIC scores 5, 6, or 7-8), we computed the PPVs with 95% confidence intervals (CIs). The PPVs were computed as the proportion of DIC diagnoses retrieved from the laboratory database (ISTH DIC score ≥5) that could be confirmed in the discharge summary or medical record (“gold standard positive”). In additional analyses, we calculated the PPVs excluding patients with a history of cancer (solid tumors and hematological cancers) and liver disease, as these conditions often are associated with coagulation abnormalities.

To examine any disparities in PPV across subgroup of patients, analyses were stratified by age (≤60 years vs >60 years), sex, year of diagnosis (2015-2016 vs 2017-2018), department (ICU vs non-ICU), etiology (eg, sepsis or malignancy), platelet count levels (<50 vs ≥50 × 109/L), fibrinogen level (<2.94 and ≥2.94 µmol/L), INR (>1.6 and ≤1.6), fibrin D-dimer (>8 and ≤8 mg/L), antithrombin activity (≤0.60 and >0.60 IU/L), aPTT (>1.5 × upper reference limit and ≤1.5 × upper reference limit), and SOFA score (0–12 and >12). In addition, we tabulated descriptive data on the patients with DIC overall, and according to ISTH score and biochemical DIC. We calculated 30-day mortality using the Kaplan-Meier method.

Data management and sampling of patients was performed using Stata version 16.0 (StataCorp, College Station, TX, USA). PPV with 95% CI was calculated using the “ci prop” function. The study was approved as a quality assurance project by all hospital directors and heads of relevant departments in accordance with the Danish Health Care Act.

3 RESULTS

An overview of the patient record selection process is provided in Figure 1. We identified the following test results from the laboratory information system database: fibrinogen, 46 832 (24 566 unique individuals); INR, 903 271 (24 566 unique individuals); platelet count, 3 224 746 (677 396 unique individuals); and fibrin D-dimer: 96 449 (57 202 unique individuals). Overall, 14 073 patients had all laboratory components of the DIC score measured on the same date. After exclusions (Figure 1) and following restriction to first-time ISTH scores, 1590 patients with an ISTH score ≥5 were eligible for inclusion. It was necessary to adjudicate 356 medical records (retrieval rate = 100%), which were selected based on a random sample, to identify the desired 225 (63.2%) patients (or 75 in each DIC score group) who had an underlying etiology potentially consistent with DIC. These 225 records were subject to further detailed medical record review. The patients were mainly hospitalized in the ICU (79%).

3.1 PPV for overt DIC

The overall PPV for an ISTH score ≥5 was 68% (95% CI, 61%-74%) for overt DIC with symptoms of bleeding and/or microthrombus formation and 83% (95% CI, 77%-87%) for overt and subclinical DIC combined (Table 2). Increasing DIC score was associated with higher PPV: 47% (95% CI, 35%-59%) for a score of 5, 68% (95% CI, 56%-78%) for a score of 6, and 88% (95% CI, 79%-94%) for a score ≥7. Further, the PPV was higher among patients admitted to an ICU than among patients at non-ICU departments (PPV = 74% vs PPV = 44%; Table 3). The PPV was also higher among patients with sepsis than among patients with malignancy (PPV = 70% vs PPV = 40%), antithrombin activity below ≤0.60 IU/L (PPV = 81% vs PPV = 49%), prolonged activated partial thromboplastin time >1.5 × upper reference limit (PPV = 80% vs PPV = 50%), and in patients with a SOFA score >12 (PPV = 81% vs PPV = 49%; Table 3). The PPVs were largely similar across age groups, sex, and calendar period. A detailed overview stratified after DIC score is showed in Table S3. The PPVs remained unchanged when excluding 83 patients with a history of cancer or liver disease (PPV overall = 67% (95% CI, 59%-75%); PPV for score 5 = 45% (95% CI, 32%-60%); PPV for score 6 = 67% (95% CI, 51%-80%); and PPV for score ≥7 = 92% (95% CI, 80%-98%).

TABLE 2. Positive predictive values for ISTH DIC score ≥5

DIC-positive cases

(overt DIC)

DIC-positive cases

(overt and subclinical DIC)

DIC-negative casesb DIC score Ratioa

PPV, %

(95% CI)

Ratioa

PPV, %

(95% CI)

No. of cases (%) Overall 152/225 68 (61-74) 186/225 83 (77-87) 39 (17) 5 35/75 47 (35-59) 56/75 75 (63-84) 19 (25) 6 51/75 68 (56-78) 60/75 80 (69-88) 15 (20) ≥7 66/75 88 (78-94) 70/75 93 (85-98) 5 (8) Abbreviations: CI, confidence interval; DIC, disseminated intravascular coagulation; PPV, positive predictive value. a Number of patients classified correct/total number of medical record reviews. b These included liver disease, hematological disease, congestive heart failure, medical treatments (eg, warfarin, chemotherapy), immune thrombocytopenia, or thrombotic thrombocytopenia purpura/hemolytic uremic syndrome. TABLE 3. Positive predictive values of ISTH DIC score for overt DIC, by patient subgroups Ratio PPV, % (95% CI) Ratio PPV, % (95% CI) Age ≤60 years >60 years 63/89 71 (60-80) 89/136 65 (57-73) Sex Male Female 87/129 67 (59-75) 65/96 68 (57-77) Time period 2015-2016 2017-2018 79/120 66 (57-74) 73/105 70 (60-78) Admission type ICU Non-ICU 131/177 74 (67-80) 21/48 44 (30-59) DIC etiology Sepsis Malignancy 96/138 70 (61-77) 4/10 40 (12-74) Platelet count <50 × 109/L ≥50 × 109/L 93/128 73 (64-80) 59/97 61 (50-71) Fibrinogen <2.94 µmol/L ≥2.94 µmol/L 28/33 85 (68-95) 124/192 65 (57-71) INR >1.6 ≤1.6 131/175 75 (68-81) 21/50 42 (28-57) Fibrin D-dimer >8.0 mg/L ≤8.0 mg/L 121/173 70 (63-77) 31/52 60 (45-73) Antithrombin ≤0.60 IU/L >0.60 IU/L 104/128 81 (73-88) 48/97 49 (39-60) aPTT >1.5 × upper reference limit a ≤1.5 × upper reference limit 55/64 86 (75-93) 37/74 50 (38-62) SOFA score ≤12 >12 46/94 49 (38–59) 105/129 81 (74–88) Abbreviations: aPTT, activated partial thromboplastin time; CI, confidence interval; DIC disseminated intravascular coagulation; INR, international normalized ratio; PPV, positive predictive value; SOFA, Sequential Organ Failure Assessment. a Restricted to individuals after July 1, 2016, due to changes in the reference interval for the analysis.

In total, 39 patients were classified as DIC negative, that is, they had a DIC score ≥5 but were judged to (i) have neither overt nor subclinical DIC based on the absence of microthrombus or bleeding symptoms described in the medical record and (ii) have one or more concurrent conditions or treatments that could explain the abnormal laboratory tests and influence the DIC score. The most prevalent conditions in the DIC negative cases were liver dysfunction (21%), hematological malignancy (33%), and medications that influenced one or more DIC score parameters (chemotherapy or vitamin K antagonists) (31%). The 39 patients who were DIC negative were similar to the patients who were overtly DIC positive regarding mean age (65 years vs 64 years), the prevalence of sepsis (85% vs 87%), and platelet count (46 × 109/L vs 42 × 109/L), but they had lower median SOFA scores (10 vs 15), higher prevalence of malignancy (41% vs 26%), higher fibrinogen (12.4 µmol/L vs 7.5 µmol/L) and higher antithrombin (0.70 IU/L vs 0.47 IU/L), lower fibrin D-dimer (9.1 mg/L vs 16.5 mg/L), and slightly lower INR (1.7 vs 2.0), and their 30-day mortality was 41% as compared to 63% in patients with overt DIC.

3.2 Characteristics of DIC patients

Among the 152 patients with overt DIC, the median age was 64 years (25th-75th percentile, 52-73 years) and 65 (43%) were women (Table 4). The most common DIC etiology was infection/sepsis (87%). Virtually all patients had confirmed or suspected thrombotic microangiopathy with a majority of the patients having signs of cold extremities, blue skin, and acute kidney injury. Venous thromboembolism was diagnosed in 14 patients (9%), while 36 (24%) experienced an arterial cardiovascular event. Overall, 115 (76%) patients experienced any bleeding, most often equivalent to WHO grade 2. The overall 30-day mortality was 63%. On the date of DIC score, patients generally had moderate to severe thrombocytopenia (median platelet count = 42 × 109/L), pronounced INR prolongation (median = 2.0), and high fibrin D-dimer (median = 16.5 mg/L); however, median fibrinogen was within the reference interval, and only 28 of the 152 overt DIC patients had fibrinogen below the cutoff <2.94 μmol/L. Low antithrombin activity was prevalent, with median (25th-75th percentile) antithrombin activity 0.47 (0.35-0.64) IU/L. The median SOFA score was 15. A positive blood culture was observed among 57 (38%) of the patients. The most frequent treatment modalities received on the day of the DIC score and on the 2 days after included inotropes (74%), blood transfusion (45%), fresh frozen plasma (51%), platelet transfusion (51%), dialysis (46%), and prophylactic low-molecular-weight heparin (42%). The use of antithrombin, surgery, fibrinogen concentrate, and tranexamic acid was less common (Table 4). A detailed overview of demographic and clinical characteristics stratified after DIC score is displayed in Table S4.

TABLE 4. Characteristics of patients with overt (n = 152) and subclinical (n = 34) DIC

Overt DIC

n = 152

Subclinical DIC

n = 34

Demographic data Age, y (median, IQR) 64 (52-73) 67 (48-77) Female sex, n (%) 65 (43) 19 (56) 30-day all-cause mortality, n (%) 95 (63) 0 (0) DIC etiology, n (%) Infection/sepsis 132 (87) 30 (88) Malignancy 40 (26) 10 (29) Othera 44 (29) 6 (18) Thrombosis, n (%) Microthrombus formation No ≤5 11 (32) Suspected 46 (30) 23 (68) Yes 101 (66) 0 (0) Venous thrombosis 14 (9) ≤5 Arterial thrombosis 36 (24) 7 (21) Bleeding No 28 (18) 25 (74) Suspected 9 (6) 2 (6) Yes (any) 115 (76) 7 (21) WHO grade 2 111 (73) 7 (21) WHO grade 3-4 47 (31) 2 (6) Laboratory results

Platelet count, ×109/L

(ref: 150-400)

42 (26-76) 66 (36-100)

Fibrinogen, μmol/L

(ref: 5.5-12.0)

7.5 (4.0-12.2) 12.6 (9.5-15.5) INR (ref: <1.2) 2.0 (1.7-2.5) 1.8 (1.3-2.0) Fibrin D-dimer, mg/Lb 16.5 (8.6-21.0) 12.6 (7.5-21.0)

Antithrombin activity, IU/L

(ref: 0.80-1.20)

0.47 (0.35-0.64) 0.70 (0.57-0.82) aPTT, sc (ref: 20-29) 49 (36-65) 33 (29-35) SAPS III score 70 (60-87) 76 (65-84) SOFA score 15 (12-18) 8 (5-14) Treatment received LMWH (prophylactic) 64 (42) 15 (44) LMWH (treatment) 13 (9) 6 (18) UFH 26 (17) ≤5 Red blood cell transfusion 69 (45) 6 (18) Platelet transfusion 78 (51) 6 (18) Fresh frozen plasma 77 (51) ≤5 Fibrinogen concentrate 16 (11) 0 (0) Tranexamic acid 11 (7) ≤5 Antithrombin 28 (18) ≤5 Surgeryd 26 (17) ≤5 Dialysis 70 (46) ≤5 ECMO 27 (18) ≤5 Inotropes 112 (74) 15 (44) Abbreviations: ALAT, alanine transaminase; aPTT, activated partial thromboplastin time; DIC, disseminated intravascular coagulation; ECMO, extracorporeal membrane oxygenation; INR, international normalized ratio; IQR, interquartile range; LMWH, low-molecular-weight heparin; ref, reference interval; SAPS, Simplified Acute Physiology Score; SOFA, Sequential Organ Failure Assessment; UFH, unfractionated heparin. a Other conditions include obstetric complications, cardiac arrest, acute pancreatitis. b Reference intervals are age specific: <55 years, <0.50 mg/L; 55-64 years, <0.60 mg/L; 65-74 years, <0.70 mg/L; 75-84 years, <0.80 mg/L; 85-94 years, <0.90 mg/L; ≥95 years, <1.0 mg/L. c Restricted to individuals after July 1, 2016, due to changes in the reference interval for the analysis. d Amputation or bowel resection. 4 DISCUSSION

Our study investigated PPV of an ISTH DIC score ≥5 for overt DIC using Danish hospital laboratory data. Overall PPV for overt (symptomatic) DIC was moderate but increased with increasing DIC score in a dose-response manner and when subclinical DIC was included. Furthermore, PPV was higher among patients with sepsis, patients with low antithrombin activity or prolonged aPTT, and patients with a high SOFA score. The main reasons for a false-positive ISTH DIC score were the presence of liver dysfunction, hematological malignancy, or pharmacological treatment that could influence either platelet count or INR (eg, vitamin K antagonist or chemotherapy).

Our study adds considerably to the sparse literature on the diagnostic a