Development of a Clinical Decision Support System for Improving the Acute Care Process of Patients Based on Clinical Variables

Master thesis in medicine

Ludvig Valentin-Askman

8 dec 2015 Background

Hemorrhagic stroke (ca 13 %)

VS.

Ischemic stroke (ca 87 %)

8 dec 2015 Background

• Stroke Subtyping Scoring Systems used for differentiation between hemorrhage and ischemia

• For prehospital triage of stroke patients?

8 dec 2015 Aim

• Provide data for development of a clinical decision support system for stroke patients through literature review

• Interviews with clinical expertise

• Construct a draft!

8 dec 2015 Results

Nine identified Stroke Subtyping Scoring Systems

differentiating between HS and IS

8 dec 2015 The Allen Score (1983)

• London, U.K.

• prospective for variable derivation, validation in same population

• 165 patients (136 IS, 29 HS, all >74 years)

• Reference test CT or autopsy

• Requires 24 h observation

8 dec 2015 The study derived eight clinical variables from a patient material consisting of 165 patients

presenting with acute stroke and consecutively recruited upon admission (14). Discriminant

coefficent values were calculated for the variables, resulting in a value that separates patients into

two groups.

Variables used in the Allen score are shown in Table 6, along with clinical features. Notably, the

score is constructed for observation of patients at least 24 hours from admission.

Table 6: Variables included in the Allen score, and their interpretation in terms of clinical features (14) Variable Clinical feature Apoplectic onset •! One!or!none!of!these! a)! Loss!of!consciousness! •! Two!or!more! b)! Headache!within!2!h! c)! Vomiting!! d)! Neck!stiffness! Level of consciousness (24 h after admission) •! Alert! •! Drowsy! •! Unconscious! Plantar response •! Both!flexor/single!extensor! •! Both!extensor! Diastolic (24 h after admission) Blood pressure in mm Hg Atheroma markers (, , •! Not!present! diabetes history) •! One!or!more! History of •! Not!present! •! Present! Previous event (TIA or stroke) •! None!! •! Any!number!of!previous!events! Heart disease •! None! •! Aortic!or!mitral!murmur! •! Cardiac!failure! •! Cardiomyopathy! •! Atrial!fibrillation! •! Cardiomegaly!(Chest!radiograph)! •! Myocardial!!within!six! months! TIA= Transient Ischemic Attack

Variable Validation

• CorrectWhen physiciansdiagnosis were inasked 148/165 to clinically patients diagnose study patients as IS or HS without using the

score, correct diagnosis was obtained in 138/165 patients, whereas the Allen score correctly8 dec 2015

diagnosed 148/165 patients (p=0.078) (14). For six patients the score indicated IS as the most likely

! 12! The Siriraj Score (1991)

• Bangkok, Thailand

• Prospective

• Acute supratentorial stroke

• Excluded subarachnoid hemorrhage, brain stem or cerebellary stroke

• 174 patients for variable derivation, 206 for validation

• Reference test CT

8 dec 2015 etiology, although CT-scan revealed HS. Four of these had small deep hematomas and two had peripheral hematomas where investigation revealed partly hemorrhagic .

! 6.1.2 The Siriraj Score

Table 7: Data for the Siriraj Score Introduction year 1991 Introduction location Bangkok, Thailand Design Prospective Included etiologies Acute supratentorial stroke Exclusion criterias Subarachnoid hemorrhage, brain stem or cerebellary stroke. Patient selection Consecutive recruitment Patient characteristics 174 for variable derivation, 206 for validation Reference test CT-scan (all patients examined within 72 hours from onset) CT= Computed Tomography

Variable Derivation

The Siriraj stroke score was proposed with the ambition to differentiate between HS and IS in supratentorial stroke syndromes (e.g. within the left or right hemisphere).

Upon admission to hospital 13 variables were recorded. The Siriraj score include five variables which were found to differ significantly between IS patients and HS patients.

Table 8: Variables and interpretation in the Siriraj score for differentation between IS and HS (15) Variable Interpretation Level of consciousness Alert= 0p, drowsy= 1p, semicoma/coma= 2p Vomiting No= 0p, yes= 1p VariableHeadache Validation within 2h No= 0p, yes= 1p Diastolic blood pressure - DuringAtheroma validation, markers the(history score of achieveddiabetes, angina, a sensitivity or forNone= HS of0p, 89.3% one or more=and sensitivity 1p for IS of intermittent claudication) 93.2%IS= Ischemic (15) .Stroke, The overall HS= Hemorrhagic predictive Stroke accuracy was 90.3% during validation, as 149/165 patients were

Formula 4: Calculation of the Siriraj Stroke Score (15) correctly diagnosed. 41 patients recieved scores between the cut off values, and were therefore not CDEDEFG&HIEJKL&HMJEL = 2.5&×&QLRLQ&JS&MJTHMDJUHTLHH + 2&×&RJVDIDTW + 2&×&ℎLFYFMℎL&ZDIℎDT&2&ℎJUEH diagnosed by the score, nor were they included in the calculation of score performance. + 0.1&×&YDFHIJQDM&]QJJY&^ELHHUEL − 3&×&FIℎLEJVF&VFEKLEH − &12

Cut off points were identified, resulting in the optimal differentiation of the IS group and the HS Table 9: Score distribution for the hemorrhagic stroke group and the ischemic stroke group in validation study (15) Score HS (n=142) IS (n=64) group.< -1 If the score is >1, HS is the most likely etiology,13 whereas a score <-1 indicates41 IS. Values >-1, <1 21 20 between>1 cut off points will remain undiagnosed. 108 3 HS= Hemorrhagic Stroke, IS= Ischemic Stroke

6.1.3 The Besson Score ! 13! 41 patients with equivocal scores! Table 10: Data for the Besson Score Introduction year 1995 8 dec 2015 Introduction Centre Hospitalier Universitaire de Grenoble, Grenoble, France location Design Prospective Included Stroke symptoms >24h, with at least unilateral motor weakness in face and/or arm and/or leg etiologies Exclusion Bilateral motor weakness, anticoagulant therapy criterias Patient selection Consecutive recruitment Patient 368 (209 men, 159 women, median 72 years, range 21-97 years, 305 IS, 63 HS) for variable characteristics derivation. 200 (122 men, 78 women, median 72 years, range 27-95 years, 167 IS, 33 HS) for validation Reference test CT-scan (within 24h) IS= Ischemic Stroke, HS= Hemorrhagic Stroke, CT= Computed Tomography

Variable Derivation

The Besson score was proposed with the aim to develop a simple SSSS for distinguishing IS from

HS with Positive Predictive Value (PPV) close to 100 %. The PPV is the probability of disease when the test indicates disease, whereas the Negative Predictive Value (NPV) is the probability of no disease when the test indicates no disease (16). The Besson score was designed for prehospital use.

The aim was not to classify all , but rather to identify patients with IS. As a result of this,

NPV has not been included in this study as an indication of score performance. Transformed IS were pooled with HS since aim was to predict intracerebral blood on CT-scan.

! 14! The Besson Score (1995)

• Grenoble, France

• Prospective

• Stroke symptoms, with at least unilat. motor weakness in face and/or arm and/or leg

• Excluded bilat. motor weakness, anticoagulant treatment

• 368 patients for variable derivation, 200 for validation

• Reference test CT

8 dec 2015

26 clinical variables were recorded. 305 patients (82.9 %) had IS and 63 patients had HS, 52 of these were cerebral hematomas and 11 were hemorrhagic infarctions.

Table 11: Besson score variables and interpretation (17) Variable Scoring Alcohol consumption 1p for everyday alcohol consumption, regardless of amount Plantar response 0p if absent, 1p if extensor ipsilateral to deficit, 2p if extensor contralateral to deficit, 3p if bilateral extensor Headache Within 2 hours before onset and/or after onset – 1p if present, 0p if absent History of hypertension 0p if absent/unknown, 1p if present History of transient neurological deficit 0p if absent/unknown, 1p if present Peripheral arterial disease 1p if history of lower limb claudication or loss of one or more ankle pulses at examination, 0p if absent History of hyperlipidemia 0p if absent/unknown, 1p if present Atrial fibrillation at admission 0p if absent, 1p if present

Formula 5: Calculation of the Besson Score (17)

• 200 patients in validation,`LHHJT &HMJEL167 IS • 72 IS correctly classified, while no HS were = 2×&aQMJℎJQ&MJTHUV^IDJT + 1.5×&bQFTIFE&ELH^JTHL + 3×&cLFYFMℎL misclassified+ 3×&cDHIJEd&JS&ℎd^LEILTHDJT − 5×&eEFTHDLTI&TLUEJQJWDMFQ&YLSDMDI − 2×&bLED^ℎLEFQ&FEILEDFQ&YDHLFHL − 1.5×&cDHIJEd&JS&ℎd^LEQD^DYLVDF − 2.5×&aIEDFQ&SD]EDQQFIDJT&JT&FYVDHHDJT

A score <1 is indicative of IS. In the prospective study all patients with a score below8 decthe threshold2015 had IS. These made up 40 % of the total of 305 patients with IS.

Variable Validation

200 patients were included in the validation study. 167 patients (83.5 %) had IS, 31 patients had cerebral hematomas (15.5 % of total cases) and two patients (1 %) had hemorrhagic infarctions.

A score below <1 was reported in 72 IS patients, correctly diagnosing 43 % of patients with IS and

36 % of all patients. The threshold did not diagnose any hemorrhages as non-hemorrhagic, resulting in a PPV at 100 % (95 % CI 93-100) for diagnosing IS.

! 15! The Greek Score (2002)

• Athens, Greece

• Prospective

• Acute stroke

• Excluded tumors, anticoagulant treatment

• 235 patients for variable derivation, 168 for validation

• Reference test CT

8 dec 2015 Table 13: Greek score patient’s characteristics and comparison IS vs. HS (18) Characteristic IS (n=192) HS (n=43) P-value Men 97 (50.5 %) 22 (51.2 %) ns Agea 69.2 ± 11.3 71.9 ± 12.4 ns History ! ! ! Atrial fibrillation 43 (22.4 %) 2 (5.1 %) <0.01 Coronary heart disease 37 (19.2%) 7 (16.5 %) ns Peripheral arterial disease 39 (20.2 %) 6 (14 %) ns Stroke 33 (17.2 %) 6 (14 %) ns TIA 28 (14.6 %) 1 (2.3 %) <0.01 Hyperlipidemia 125 (65.1 %) 19 (44.2 %) ns Alcohol consumption 81 (42.2 %) 20 (46.5 %) ns Contraceptive pill use 0 0 ! Hypertension 101 (52.6 %) 26 (60.5 %) ns Diabetes mellitus 32 (16.7 %) 6 (13.9 %) ns Cigarette smoking 74 (38.5 %) 18 (41.8 %) ns Clinical evaluation ! ! ! Acute onset of deficit 44 (22.9 %) 20 (46.5 %) <0.001 Headache 49 (25.5 %) 19 (44.2 %) <0.001 Vomiting 11 (5.7 %) 30 (69.8 %) <0.001 Systolic blood pressure (mm Hg)a 161.7 ± 21.5 165.4 ± 15.3 ns Diastolic blood pressure (mm Hg)a 96.5 ± 11.5 100.5 ± 8.6 ns Glascow coma scalea 10.1 ± 2.8 10.9 ± 1.9 ns Pupil derangement 6 (3.1 %) 1 (2.3 %) ns Abnormal plantar response 154 (80.2 %) 36 (83.7 %) ns Kernig and/or Brudzinski sign 3 (1.6 %) 1 (2.3 %) ns Neck stiffness 36 (18.8 %) 17 (39.5 %) <0.001 Decreased level of consciousness 48 (25 %) 38 (88.4 %) <0.001 Neurological deterioration within first 3 h 24 (12.5 %) 35 (81.4 %) <0.001 Laboratory data ! ! ! Atrial fibrillation in ECG on admission 52 (27.1 %) 3 (7 %) <0.01 Platelet counta 362 150 ± 94 960 359 220 ± 79 880 ns Mean volume (fl)a 11.1 ± 3.2 9.9 ± 2.7 ns White blood cell count (L-1)a 6920 ± 1180 14 880 ± 1360 <0.001 Plasma fibrinogen (mg dL-1)a 460 ± 170 730 ± 160 <0.001 Plasma protrombin time (s)a 14.2 ± 2.1 13.9 ± 1.9 ns IS= Ischemic Stroke, HS= Hemorrhagic Stroke, TIA= Transient Ischemic Attack, ECG= Electrocardiogram aMean±SD, ns= not significant

During variable derivation, 43 patients (18.3 %) had HS (18). Previous medical history of atrial fibrillation, previous TIA, and atrial fibrillation at admission were found significantly more frequent in IS patients. Acute onset of deficit, headache, vomiting, neck stiffness, decreased level of consciousness, neurological deterioration within 3 hours from admission, fibrinogen >500 mg dL-1, and white blood cell count (WBC) > 12 000 mL-1 had higher prevalence in the HS group.

Formula 6: Calculation of the Greek Score (18) fELLK&CMJEL = 6&×& TLUEJQJWDMFQ&YLILEDJEFIDJT&ZDIℎDT&3&ℎJUEH&SEJV&FYVDHHDJT + 4&×& RJVDIDTW + &4&×& i`j > 12&000 + 3&×&(YLMELFHLY&QLRLQ&JS&MJTHMDJUHTLHH) WBC= White Blood Cell Count

! 17!

• 168 patients in validation • 126 patients classified as IS, only 1 had HS • 13 patients with equivocal scores

8 dec 2015 The Kurashiki Prehospital Stroke Subtyping Score (2011)

• Kurashiki, Japan

• Retrospective

• Excluded trauma, tumor, TIA, subarachnoid hemorrhage

• 227 patients (127 IS, 100 HS)

• Reference test CT or MRI

8 dec 2015 6.1.5 The Kurashiki Prehospital Stroke Subtyping Score (KP3S)

Table 15: Data for KP3S Introduction year 2011 Introduction location Kurashiki, Japan Design Retrospective Included etiologies Acute stroke Exclusion criterias Cerebral hemorrhage from trauma or brain tumor, TIA, subarachnoid hemorrhage Patient selection Consecutive patients with IS or HS transferred by emergency medical staff Patient characteristics 227 (median age 71 years, 139 males, 127 IS, 100 HS) Reference test CT- or MRI-scan TIA= Transient Ischemic Attack, IS= Ischemic Stroke, HS= Hemorrhagic Stroke, CT= Computed Tomography, MRI= Magnetic Resonance Imaging

Variable Derivation

The KP3S was developed with the aim distinguishing IS from HS in a prehospital setting. If the score could provide prehospital medical staff with the probability of IS, the staff could bring the patient to a hospital with resources to start thrombolytic treatment.

Clinical characteristics were compared between the two patient groups. Presence of atrial fibrillation, lack of disturbance of consciousness, and diastolic blood pressure <100 mm Hg were found to be independent factors associated with IS.

Table 16: KP3S variables and distribution within IS and HS patient groups. Percentage in parentheses (19) Variable IS (n=127) HS (n=100) p-value Atrial fibrillation 51 (40) 4 (4) <0.001 Mean DBP 90 mm Hg 102 mm Hg <0.001 Lack of disturbance of 69 (54) 30 (30) <0.001 consciousness IS= Ischemic Stroke, HS= Hemorrhagic Stroke, DBP= Diastolic Blood Pressure

Formula 7: Calculation of the Kurashiki Prehospital Stroke Sbutyping Score, KP3S (19) nb3C = 2&×& ^ELHLTML&JS&FIEDFQ&SD]EDQQFIDJT + YDFHIJQDM&]QJJY&^ELHHUEL < 100&VV&cW + &&&&(QFMK&JS&YDHIUE]FTML&JS&MJTHMDJUHTLHH)

Variable Validation 100% Analysis of ROC90% -curves estimated the appropriate cut off values for IS. When the proposed score 80% was applied to70% the study population the prevalence of IS was found to increase with higher scores. 60% 50% Hemorrhagic!Stroke 40% Ischemic!Stroke 30% ! 20% 19! 10% 0% 0!points 1!point 2!points 3!points

Fig. 2: Distribution of Hemorrhagic and Ischemic stroke for each possible score in KP3S 8 dec 2015

If the KP3S generated a score >1 point, the sensitivity for IS was 64 %, specificity 85 %, PPV 84 %,

and NPV 65 %.

! 6.1.6 Score proposed by Stürmer et al.

Table 17: Data for score proposed by Stürmer et al. Introduction year 2002 Introduction University Hospital of Ulm, Germany location Design Retrospective Included IS, TIA, HS, subarachnoid hemorrhage etiologies Exclusion Patients missing continuous variables (blood pressure and heart rate) criterias Patient selection All patients with discharge diagnosis of IS, TIA, HS, or subarachnoid hemorrhage admitted during a period of 30 months. Patient 540, divided into 2 datasets by random number: one for variable derivation and one for characteristics validation. 98 (18.1 %) with hemorrhage and 442 (81.9 %) with ischemia, mean age 70.3 years, 53.7 % women. Reference test Clinical data and CT-scan IS= Ischemic Stroke, TIA= Transient Ischemic Attack, HS= Hemorrhagic Stroke, CT= Computed Tomography

Variable Derivation

Stürmer et al. proposed a SSSS suitable for prehospital use, to increase channeling of patients with

possible IS to hospitals with resources to start thrombolytic treatment.

! 20! Score proposed by Stürmer et al. (2002)

• Ulm, Germany

• Retrospective

• IS, TIA, HS, Subarachnoid hemorrhage

• 540 patients, divided into two datasets; one for variable derivation and one for validation

• Reference test clinical data & CT

8 dec 2015 Diagnoses were pooled as either ischemic, IS or TIA, or hemorrhage, subarachnoid hemorrhage or

HS. Differential diagnostic Odds Ratios (OR) were calculated for each variable, which quantifies how well each variable perform in distinguishing ischemia from hemorrhage. Comparison with the

Allen, the Siriraj, and the Besson score, was done with original cut off values. The proposed score consists of these variables:

•! Age •! Obesity (Body Mass Index>27.3 for women, >27.8 for men. If height was missing: obesity if >90kg) •! Anamnestic stroke/TIA •! Peripheral arterial disease (history of or at least unilateral lack of ankle pulse) •! Onset during physical activity •! Headache within 2 hours from onset •! Impaired consciousness •! Hemisyndrome (hemiplegia/hemiparesis) •! Meningismus •! Systolic blood pressure

98 (18.1 %) patients had hemorrhage, remaining 442 patients (81.9 %) had ischemia (20). Almost half of patients had history of cerebral event, 26.6% had history of stroke, and 19.3% had history of

TIA. Increasing age was linked to lower probability of hemorrhage. OR’s were calculated, OR <1.0 indicate association with ischemia and OR >1.0 indicate association with hemorrhage.

Table 18: Variables included in score proposed by Stürmer, presented with Odds Ratios (20) Variable Odds Ratio (95 % CI) Debut of symptoms between 6-9AM or at wake up 0.53 (0.22-1.26) Debut of symptoms during rest(sleeping, lying, or 0.45 (0.17-1.16) sitting) Headache within 2 hours from onset 7.66 (3.30-17.8) Vomiting 4.97 (2.29-10.8) Impaired consciousness 3.75 (1.96-7.18) Meningismus 10.3 (2.96-35.7) Systolic blood pressure ≥200 mm Hg 3.35 (1.62-6.93) Hemiplegia/hemiparesis 0.24 (0.12-0.47) Obesity 0.46 (0.12-0.69) History of stroke/TIA 0.28 (0.12-0.61) Peripheral arterial disease 0.30 (0.10-0.89) CI= Confidence Interval, TIA= Transient Ischemic Attack

• 270 patients in validation ! 21! • 67 patients classified as IS, only one had HS • = 24.8 % of IS patients were classified

8 dec 2015 The Stroke Data Bank Score (2002)

• Multicenter, North-East U.S.

• Prospective for variable derivation, cohort material for validation

• Acute stroke

• Excluded TIA, trauma, tumor, hematological disease, subarachnoid hemorrhage

• 1510 patients (1273 IS, 237 HS)

• Reference test CT 8 dec 2015 A score was based on clinical variables, with score range from -10 to +10. Analysis of ROC-curves when applied to the SDB material generated an optimal cut off at +2.0. For scores ≤2.0 points, HS is most likely diagnosis and if score >2.0 points, IS is most likely diagnosis. When using this cut off,

76 % of HS could be classified correctly, whereas 17 % of IS were misclassified as HS. This resulted in a sensitivity of 76 %, a specificity of 83 %, PPV of 46 % and NPV of 95 % for HS.

Variable Validation

During validation 68 % of HS patients were correctly classified and 24 % of IS patients were misclassified as HS. The Siriraj score generated a sensitivity of 24 % and a specificity of 97.4 %. Out of 58 included patients, the Siriraj score correctly classified 14 cases.

Table 23: Calculation of the Stroke Data Bank Score (22) ! No Yes Unknown Patient > 55 years old 0 +1.0 U Patient is a man 0 -1.0 ! Has patient ever had history of: ! ! ! TIA or stroke 0 +2.0 U Angina 0 +2.0 U Diabetes 0 +1.5 U At the time of onset was there: ! ! ! Focal deficit 0 +2.5 U Deficit presented on awakening 0 +1.0 U Patient is not alert 0 -3.0 U Severe headache 0 -2.0 U Vomiting 0 -1.5 U Initial blood pressure > 200/120 mm Hg 0 -1.5 ! More than one of previous questions answered unknown 0 -1.0 ! Rating -10 to + 2 HS ! >2 to +10 IS TIA= Transient Ischemic Attack, HS= Hemorrhagic Stroke, IS= Ischemic Stroke

• 68 % of HS patients correctly classified ! • ! 24 % of IS patients misclassified as HS ! ! ! ! ! 8 dec 2015

! 26! Score proposed by Woisetschläger et al. (2002)

• Vienna, Austria

• Retrospective

• 224 patients (118 HS, 106 IS)

• Reference test CT

8 dec 2015

Table 25: Distribution of leading presenting symptom, HS vs. IS (23) ! HS IS P-value Number of patients 118 106 ! Hemisymptoms 14 (15.9 %) 74 (84.1 %) <0.0001 TableImpaired 25: Distribution level of consciousness of leading presenting symptom,70 (94.6 HS %) vs. IS (23) 4 (5.4 %) <0.0001 Aphasia! HS14 (63.6 %) IS8 (36.4 %) P0.22-value CephaleaNumber of patients 11811 (67.4 %) 1066 (35.3 %) !0.25 VomitingHemisymptoms 142 (22.2 (15.9 %) %) 747 (77.8(84.1 %)%) <0.00010.19 SeizureImpaired level of consciousness 704 (44.4 (94.6 %) %) 45 (5.4 (55.6 %) %) <0.00010.73 SyncopeAphasia 143 (60.0 (63.6 %) %) 82 (36.4 (40.0 %) %) 0.220.69 HS=Cephalea Hemorrhagic Stroke, IS= Ischemic Stroke 11 (67.4 %) 6 (35.3 %) 0.25 Vomiting 2 (22.2 %) 7 (77.8 %) 0.19 TableSeizure 26: Calculation of score proposed by Woisetschläger4 (44.4 %) et al. (23) 5 (55.6 %) 0.73 !Syncope 3 (60.0 %) Score 2 (40.0 %) 0.69 HS= Hemorrhagic Stroke, IS= Ischemic Stroke Evaluation of presenting symptom ! Impaired level of consciousness 3 TableHemisymptoms 26: Calculation of score proposed by Woisetschläger et al. (23)-1 Other! Score1 Evaluation of presentingmedical history symptom ! ! HistoryImpaired of level diabetes of consciousness 3-1 HistoryHemisymptoms of hypertension --11 Other 1 Evaluation of medical history ! History of diabetes -1 VariableHistory of Validation hypertension -1

Table 27: Distribution of HS and IS for each possible score (23) VariableScore Validation HS IS -3 0 7 -2 4 40 Table 27: Distribution of HS and IS for each possible score (23) -1 12 32 Score HS IS 0 21 20 -3 0 7 1 13 3 -2 4 40 2 30 4 -1 12 32 3 38 0 0 21 20 HS= Hemorrhagic Stroke, IS= Ischemic Stroke 1 13 3 2 30 4 Due3 to equivocal score38s 41 patients (18.3 %)0 remained unclassified.8 dec 2015 By analysis of the AUC in HS= Hemorrhagic Stroke, IS= Ischemic Stroke ROC -curves, the accuracy of the score was assessed. AUC was found to be 0.90 (95 % CI 0.86- Due to equivocal scores 41 patients (18.3 %) remained unclassified. By analysis of the AUC in 0.94). ROC-curves, the accuracy of the score was assessed. AUC was found to be 0.90 (95 % CI 0.86-

0.94).

! 28!

! 28! Score proposed by Runchey et al. (2010)

• Systematic review over 19 prospective studies (1966-2010)

• Excluded subarachnoid hemorrhage

• 6438 patients (1528 HS, 4910 IS)

• Reference test CT or autopsy

8 dec 2015 Score proposed by Runchey et al. (2010)

• Included the Besson & the Siriraj Score

• Greek Score & Score proposed by Woisetschläger excluded due to lack of validation beyond original study

• Allen score excluded due to requirement of data 24h after presentation

• Many variables!

8 dec 2015 6.2 Summary of Stroke Subtyping Scoring Systems

27 variables differentiating between IS and HS have been identified from previously proposed SSSS.

Table 28: Variables differentiating IS from HS, found within each Stroke Subtyping Scoring System

Variable Allen Siriraj Besson Greek KP3S Stürmer Runchey SDBS Woisetsch läger Level of conscioussness X X ! X X X X X X Headache X X X ! ! X X X ! Vomiting X X ! X ! X X X ! Neck stiffness/meningismus X ! ! ! ! X X ! ! Plantar response X ! X ! ! ! X ! ! Blood pressure (systolic and/or diastolic) X X ! ! X X X X ! Atheroma markers(coronary X X X ! ! X X X X disease/claudication/obesity/diabetes/peripheral arterial disease/hyperlipidemia) History of hypertension X ! X ! ! ! X ! X Previous TIA/stroke X ! X ! ! X X X ! Heart disease X ! ! ! ! ! ! ! ! Hemisymptoms ! ! ! ! ! X X ! X Alcohol consumption ! ! X ! ! ! X ! ! Age ! ! ! ! ! ! X X ! Atrial fibrillation(previous/present) ! ! X ! X ! X ! ! Progressing neurological deterioration ! ! ! X ! ! ! ! ! White blood cell count >12 000 mL-1 ! ! ! X ! ! ! ! ! Debut of symptoms at wake-up/morning ! ! ! ! ! X ! X ! Debut of symptoms during rest ! ! ! ! ! X ! ! ! Seizures at onset ! ! ! ! ! ! X ! ! Cervical ! ! ! ! ! ! X ! ! Focal deficit at onset ! ! ! ! ! ! ! X ! Gender ! ! ! ! ! ! X X ! Cigarette smoking ! ! ! ! ! ! X ! ! Acute onset of deficit ! ! ! ! ! ! X ! ! Kernig sign, Brudzinski sign, or both ! ! ! ! ! ! X ! ! Xanthocromia in cerebrospinal fluid ! ! ! ! ! ! X ! ! Clinician’s overall impression ! ! ! ! ! ! X ! ! TIA= Transient Ischemic Attack 6.3 Validation of Variables

During the project it became evident most variables were not available in the Ambulink or RETTS journal systems in a form useful for validation. Variables might be available in free text form, but this was considered to be of no use for retrospective validation. The only variables available from previous Medfield studies were blood pressure and age as they were obtained during study C, and patients were included as either Hemorrhage (HS), Infarction (IS), or as undergoing thrombolytic

! 29! 6.4 Construction of CDSS

6.4.1 Construction of HS risk calculator

Based on data from Runchey et al., a HS risk calculator was constructed, using variables assessable in a prehospitalSelected setting. variables

Table 33: Variables with Positive Likelihood ratio and Negative Likelihood ratio used in the CDSS for prehospital assessment of risk of hemorrhagic stroke ! LR+ (95% CI) LR- (95% CI) Risk factors ! ! Age ≤60 years 1.7 (1.4-1.9) 0.71 (0.63-0.82) Previous stroke 0.59 (0.17-2.0) 1.1 (0.88-1.4) Previous TIA 0.34 (0.18-0.65) 1.2 (1-1-1.3) Atrial fibrillation 0.44 (0.25-0.78) 1.1 (1.05-1.1) Peripheral artery disease 0.41 (0.2-0.83) 1.1 (1.0-1.1) 0.44 (0.31-0.61) 1.1 (1.0-1.3) (/Angina) Hyperlipidemia 0.48 (0.2-1.1) 1.1 (1.1-1.1) Symptoms ! ! Seizures 4.7 (1.6-14) 0.93 (0.9-0.96) Vomiting 3.0 (1.7-5.5) 0.73 (0.59-0.91) Headache 2.9 (1.7-4.8) 0.66 (0.56-0.77) Loss of consciousness 2.6 (1.6-4.2) 0.65 (0.52-0.82) Physical signs ! ! LOC: Coma (RLS ≥4) 6.2 (3.2-12) - LOC: Drowsy(RLS 2-3) 2.0 (1.0-3.9) - LOC: Alert (RLS 1) 0.35 (0.24-0.5) - Neck stiffness 5.0 (1.9-12.8) 0.83 (0.75-0.92) DBP >110 mm Hg 4.3 (1.4-14) 0.59 (0.39-0.89) LR= Likelihood Ratio, CI= Confidence Interval, TIA= Transient Ischemic Attack, LOC= Level Of Consciousness, RLS= Reaction Level Scale, DBP= Diastolic Blood Pressure 8 dec 2015

Variable!1 Variable!2 Variable!3

Fig. 4: Schedule for probability calculations within the HS risk calculator

The HS risk calculator will use each variable answered with Yes or No, and present the posttest probability. Each new probability will then be used as pretest probability for the next applicable variable. Variables can also be answered as unknown; the calculator will then skip calculations for that variable.

! 32! % of true-positives LR+ = % of false-positives % of false-negatives LR- = % of true-negatives

8 dec 2015 8 dec 2015 8 dec 2015 8 dec 2015 Example of Use

• 56 year old male with suspected stroke

• No prior significant medical history

• Drowsy

• Vomits once after pick-up

• BP 196/112 mm Hg

2 Thoughts…

• Most variables unavailable for retrospective validation

• All variables will not available in a clinical setting

• Proof-of-concept

• Current study aimed at validating variables

8 dec 2015