Evaluating automated machine learning platforms for use in healthcare.

Objective: To describe development and application of a checklist of criteria for selecting an automated machine learning (Auto ML) platform for use in creating clinical ML models. Materials and Methods: Evaluation criteria for selecting an Auto ML platform suited to ML needs of a local health district were developed in 3 steps: (1) identification of key requirements, (2) a market scan, and (3) an assessment process with desired outcomes. Results: The final checklist comprising 21 functional and 6 non-functional criteria was applied to vendor submissions in selecting a platform for creating a ML heparin dosing model as a use case. Discussion: A team of clinicians, data scientists, and key stakeholders developed a checklist which can be adapted to ML needs of healthcare organizations, the use case providing a relevant example. Conclusion: An evaluative checklist was developed for selecting Auto ML platforms which requires validation in larger multi-site studies.

3D exoscopic versus microscopic superficial temporal artery to middle cerebral artery bypass surgery for moyamoya disease - a comparative series.

PURPOSE: Superficial temporal artery to middle cerebral artery (STA-MCA) direct bypass surgery is the most common surgical procedure to treat moyamoya disease (MMD). Here, we aim to compare the performance of the 3D exoscope in bypass surgery with the gold standard operative microscope. METHODS: All direct STA-MCA bypass procedures performed at a single university hospital for MMD between 2015 and 2023 were considered for inclusion. Data were retrospectively collected from patient files and surgical video material. From 2020 onwards, bypass procedures were exclusively performed using a digital three-dimensional exoscope as visualization device. Results were compared with a microsurgical bypass control group (2015-2019). The primary endpoint was defined as total duration of surgery, duration of completing the vascular anastomosis (ischemia time), bypass patency, number of stiches to perform the anastomosis, added stiches after leakage testing of the anastomosis and the Glasgow outcome scale (GOS) at last follow-up as secondary outcome parameter. RESULTS: A total of 16 consecutive moyamoya patients underwent 21 STA-MCA bypass procedures. Thereof, six patients were operated using a microscope and ten patients using an exoscope (ORBEYE® n = 1; AEOS® n = 9). Total duration of surgery was comparable between devices (microscope: 313 min. ± 116 vs. exoscope: 279 min. ± 42; p = 0.647). Ischemia time also proved similar between groups (microscope: 43 min. ± 19 vs. exoscope: 41 min. ± 7; p = 0.701). No differences were noted in bypass patency rates. The number of stiches per anastomosis was similar between visualization devices (microscope: 17 ± 4 vs. exoscope: 17 ± 2; p = 0.887). In contrast, more additional stiches were needed in microscopic anastomoses after leakage testing the bypass (p = 0.035). CONCLUSION: Taking into account the small sample size, end-to-side bypass surgery for moyamoya disease using a foot switch-operated 3D exoscope was not associated with more complications and led to comparable clinical and radiological results as microscopic bypass surgery.

The adverse inpatient medication event and frailty (AIME-frail) risk prediction model.

BACKGROUND: Medication harm affects between 5 and 15% of hospitalised patients, with approximately half of the harm events considered preventable through timely intervention. The Adverse Inpatient Medication Event (AIME) risk prediction model was previously developed to guide a systematic approach to patient prioritisation for targeted clinician review, but frailty was not tested as a candidate predictor variable. AIM: To evaluate the predictive performance of an updated AIME model, incorporating a measure of frailty, when applied to a new multisite cohort of hospitalised adult inpatients. METHODS: A retrospective cohort study was conducted at two tertiary Australian hospitals on patients discharged between 1st January and April 31, 2020. Data were extracted from electronic medical records (EMRs) and clinical coding databases. Medication harm was identified using ICD-10 Y-codes and confirmed by senior pharmacist review of medical records. The Hospital Frailty Risk Score (HFRS) was calculated for each patient. Logistic regression analysis was used to construct a modified AIME model. Candidate variables of the original AIME model, together with new variables including HFRS were tested. Performance of the final model was reported using area under the curve (AUC) and decision curve analysis (DCA). RESULTS: A total of 4089 patient admissions were included, with a mean age ± standard deviation (SD) of 64 years (±19 years), 2050 patients (50%) were males, and mean HFRS was 6.2 (±5.9). 184 patients (4.5%) experienced one or more medication harm events during hospitalisation. The new AIME-Frail risk model incorporated 5 of the original variables: length of stay (LOS), anti-psychotics, antiarrhythmics, immunosuppressants, and INR greater than 3, as well as 5 new variables: HFRS, anticoagulants, antibiotics, insulin, and opioid use. The AUC was 0.79 (95% CI: 0.76-0.83) which was superior to the original model (AUC = 0.70, 95% CI: 0.65-0.74) with a sensitivity of 69%, specificity of 81%, positive predictive value of 0.14 (95% CI: 0.10-0.17) and negative predictive value of 0.98 (95% CI: 0.97-0.99). The DCA identified the model as having potential clinical utility between the probability thresholds of 0.05-0.4. CONCLUSION: The inclusion of a frailty measure improved the predictive performance of the AIME model. Screening inpatients using the AIME-Frail tool could identify more patients at high-risk of medication harm who warrant timely clinician review.

Changes in treatment of intracranial aneurysms during the last decade in a large European neurovascular center.

OBJECTIVE: Treatment modality for ruptured and unruptured intracranial aneurysms has shifted during the last two decades from microsurgical treatment towards endovascular treatment. We present how this transition happened in a large European neurovascular center. METHODS: We conducted a retrospective observational study consecutive patients treated for an unruptured or ruptured intracranial aneurysm at Helsinki University Hospital during 2012-2022. We used Poisson regression analysis to report age-adjusted treatment trends by aneurysm location and rupture status. RESULTS: A total of 2491 patients with intracranial aneurysms were treated (44% ruptured, 56% unruptured): 1421 (57%) surgically and 1070 (43%) endovascularly. A general trend towards fewer treated aneurysms was noted. The proportion of patients treated surgically decreased from 90% in 2012 to 20% in 2022. The age-adjusted decrease of surgical versus endovascular treatment was 6.9%/year for all aneurysms, 6.8% for ruptured aneurysms, and 6.8% for unruptured aneurysms. The decrease of surgical treatment was most evident in unruptured vertebrobasilar aneurysms (10.8%/year), unruptured communicating artery aneurysms (10.1%/year), ruptured communicating artery aneurysms (10.0%/year), and ruptured internal carotid aneurysms (9.0%/year). There was no change in treatment modality for middle cerebral artery aneurysms, of which 85% were still surgically treated in 2022. A trend towards an increasing size for treated ruptured aneurysms was found (p = 0.033). CONCLUSION: A significant shift of the treatment modality from surgical to endovascular treatment occurred for all aneurysm locations except for middle cerebral artery aneurysms. Whether this shift has affected long-term safety and patient outcomes should be assessed in the future.

Even short-term training improves the skills of novice exoscope users: a prospective laboratory experiment.

BACKGROUND: The surgical 3D exoscopes have recently been introduced as an alternative to the surgical microscopes in microneurosurgery. Since the exoscope availability is still limited, it is relevant to know whether even a short-term exoscope training develops the skills needed for performing exoscope-assisted surgeries. METHODS: Ten participants (six consultants, four residents) performed two laboratory bypass test tasks with a 3D exoscope (Aesculap Aeos®). Six training sessions (6 h) were performed in between (interval of 2-5 weeks) on artificial models. The participants were divided into two groups: test group (n = 6) trained with the exoscope and control group (n = 4) with a surgical microscope. The test task was an artificial end-to-side microsurgical anastomosis model, using 12 interrupted 9-0 sutures and recorded on video. We compared the individual as well as group performance among the test subjects based on suturing time, anastomosis quality, and manual dexterity. RESULTS: Altogether, 20 bypass tasks were performed (baseline n = 10, follow-up n = 10). The median duration decreased by 28 min and 44% in the exoscope training group. The decrease was steeper (29 min, 45%) among the participants with less than 6 years of microneurosurgery experience compared to the more experienced participants (13 min, 24%). After training, the participants with at least 1-year experience of using the exoscope did not improve their task duration. The training with the exoscope led to a greater time reduction than the training with the microscope (44% vs 17%). CONCLUSIONS: Even short-term training with the exoscope led to marked improvements in exoscope-assisted bypass suturing among novice microneurosurgeons. For the more experienced participants, a plateau in the initial learning curve was reached quickly. A much longer-term effort might be needed to witness further improvement in this user group.

Artificial intelligence in medical referrals triage based on Clinical Prioritization Criteria.

The clinical prioritisation criteria (CPC) are a clinical decision support tool that ensures patients referred for public specialist outpatient services to Queensland Health are assessed according to their clinical urgency. Medical referrals are manually triaged and prioritised into three categories by the associated health service before appointments are booked. We have developed a method using artificial intelligence to automate the process of categorizing medical referrals based on clinical prioritization criteria (CPC) guidelines. Using machine learning techniques, we have created a tool that can assist clinicians in sorting through the substantial number of referrals they receive each year, leading to more efficient use of clinical specialists' time and improved access to healthcare for patients. Our research included analyzing 17,378 ENT referrals from two hospitals in Queensland between 2019 and 2022. Our results show a level of agreement between referral categories and generated predictions of 53.8%.

Anterior inferior cerebellar artery (AICA) aneurysms: a radiological study of 15 consecutive patients.

Introduction: The aneurysms of the anterior inferior cerebellar artery (AICA) are rare lesions of the posterior circulation and to treat them is challenging. We aim to present anatomical and morphological characteristics of AICA aneurysms in a series of 15 patients. Method: The DSA and CT angiography images of AICA aneurysms in 15 consecutive patients were analyzed retrospectively. Different anatomical characteristics were quantified, including morphology, location, width, neck width, length, bottleneck factor, and aspect ratio. Results: Eighty percent of the patients were females. The age was 52.4 ± 9.6 (mean ± SD) years. 11 patients were smokers. Ten patients had a saccular aneurysm and five patients had a fusiform aneurysm. Aneurysm in 10 patients were located in the proximal segment, in three patients in the meatal segment, and in two patients in the distal segment. Ten out of 15 patients presented with a ruptured aneurysm. The size of AICA aneurysms was 14.8 ± 18.9 mm (mean ± SD). The aspect ratio was 0.92 ± 0.47 (mean ± SD) and bottleneck factor was 1.66 ± 1.65 (mean ± SD). Conclusion: AICA aneurysms are rare lesions of posterior circulation predominantly found in females, present predominantly with subarachnoid hemorrhage, and are mostly large in size.

Evaluation of enamel roughness after orthodontic debonding and clean-up procedures using zirconia, tungsten carbide, and white stone burs: an in vitro study.

BACKGROUND: The main goal of orthodontic debonding is to restore the enamel surface as closely as possible to its pretreatment condition without iatrogenic damage. This study aimed to compare the effects of different adhesive removal burs; zirconia burs, tungsten carbide burs, and white stone burs on enamel surface roughness. MATERIALS AND METHODS: Total sample of 72 extracted premolars was randomly divided into three equal groups (n = 24) depending on the method of adhesive removal: zirconia burs (ZB); tungsten carbide burs (TC); and white stones (WS). The metal brackets were bonded using Transbond XT orthodontic adhesive (3 M Unitek, Monrovia, CA, USA) and debonded after 24 h using a debonding plier, then the ARI was assessed. The adhesive remnants were removed using the different burs and Final polishing was performed using Sof-lex discs and spirals. Thirteen samples from each group were evaluated using a Mitutoyo SJ-210 profilometer to determine average surface roughness (Ra) and three samples from each group were examined under Scanning Electron Microscopy (SEM) to determine EDI score. The evaluations were performed at three time points; before bonding (T0), after adhesive removal (T1) and after polishing (T2) and the time consumed for adhesive removal by burs was recorded in seconds. The data were analyzed statistically by ANOVA, Tukey's test and Kruskal-Wallis H-test. RESULTS: Kruskal-Wallis H-test showed no statistically significant difference of ARI in all studied groups (p = 0.845) and two-way mixed ANOVA revealed that all burs significantly increased surface roughness at T1 compared to T0 (p < 0.001) in all groups with the lowest Ra values were observed in the ZB group, followed by the TC group, and WS group. The fastest procedure was performed with WS, followed by ZB, then TC bur (p < 0.001). After polishing (T2), Ra values showed no significant difference in ZB group (P = 0.428) and TC group (P = 1.000) as compared to T0, while it was significant in WS group (p < 0.001). CONCLUSION: zirconia bur was comparable to tungsten carbide bur and can be considered as alternative to white stone which caused severe enamel damage. The polishing step created smoother surface regardless of the bur used for resin removal.

Dentofacial effects of miniscrew-anchored maxillary protraction on prepubertal children with maxillary deficiency: a randomized controlled trial.

BACKGROUND: The introduction of bone-anchored maxillary protraction eliminated the side effects of facemask in the early treatment of patients with maxillary retrusion. This study aimed to evaluate the effects of miniscrew-anchored maxillary protraction (MAMP) and compare them with the growth changes in an untreated control group in growing patients with Class III malocclusion. METHODS: Forty growing patients with Class III malocclusion and retrognathic maxilla were randomly allocated into two groups: treated and control groups. In the treated group, patients were treated with full-time intermaxillary Class III elastics (C3E) anchored by a hybrid hyrax (HH) in the maxilla and a bone-supported bar in the mandible. Protraction was stopped after obtaining a positive overjet. Cephalometric radiographs were acquired before and after the treatment. Data were statistically analyzed on an intention-to-treat basis. Intergroup comparisons were also made using analysis of covariance with the readings at T0 as a covariate. RESULTS: Forty patients agreed to participate, and 30 of them completed the study (treated group, n = 17; control group, n = 13). The average treatment duration was 11.9 months. MAMP resulted in a significant maxillary advancement (A-VR, 4.34 mm) with significant control over the mandibular growth. No significant increase in the mandibular plane angle was found in the treated group compared with the control group. The upper and lower incisors showed significant protrusion in the treated group. CONCLUSIONS: Within the limitations of this study and high attrition rate, the MAMP protocol can effectively increase maxillary forward growth with good control over the growth of the mandible antero-posteriorly and vertically.

Taming the exoscope: a one-year prospective laboratory training study.

PURPOSE: Digital 3D exoscopes have been recently introduced as an alternative to a surgical microscope in microneurosurgery. We designed a laboratory training program to facilitate and measure the transition from microscope to exoscope. Our aim was to observe the effect of a one-year active training on microsurgical skills with the exoscope by repeating a standardized test task at several time points during the training program. METHODS: Two board-certified neurosurgeons with no previous exoscope experience performed the same test tasks in February, July, and November during a 12-month period. In between the test tasks, both participants worked with the exoscope in the laboratory and assisted during clinical surgeries on daily basis. Each of the test segments consisted of repeating the same task 10 times during one week. Altogether, 60 test tasks were performed, 30 each. The test task consisted of dissecting and harvesting the ulnar and radial arteries of the second segment of a chicken wing using an exoscope (Aesculap AEOS). Each dissection was recorded on video and analyzed by two independent evaluators. We measured the time required to complete the task as well as several metrics for evaluating the manual skills of the dissection and handling of the exoscope system. RESULT: There was a clear reduction in dissection time between the first and the last session, mean 34 min (SD 5.96) vs. 26 min (SD 8.69), respectively. At the end of the training, both neurosurgeons used the exoscope more efficiently utilizing more available options of the device. There was correlation between the dissection time and several of the factors we used for evaluating the work flow: staying in focus, zoom control, reduction of unnecessary movements or repetitive manual motions, manipulation technique of the vessel under dissection, handling of the instruments, and using them for multiple dissection purposes (stretching, cutting, and splitting). CONCLUSION: Continuous, dedicated long-term training program is effective for microsurgical skill development when switching from a microscope to an exoscope. With practice, the micromotor movements become more efficient and the use of microinstruments more versatile.

Recurrence of brain arteriovenous malformations in pediatric patients: a long-term follow-up study.

BACKGROUND: Previously thought to be congenital, AVMs have shown evidence of de-novo formation and continued growth, thus shifting thoughts on their pathophysiology. Pediatric AVM patients have been reported to be more prone to develop AVM recurrence after a seemingly complete cure. Therefore, we assessed the risk of AVM treated in childhood to recur in adulthood after a long-term follow-up in our own cohort. METHODS: Control DS-angiography was arranged during 2021-2022 as part of a new protocol for all AVM patients who were under 21 years of age at the time of their treatment and in whom the treatment had occurred at least five years earlier. Angiography was offered only to patients under 50 years of age at the time of the new protocol. The complete eradication of AVM after the primary treatment had been originally confirmed with DSA in every patient. RESULTS: A total of 42 patients participated in the late DSA control, and 41 of them were included in this analysis after excluding the patient diagnosed with HHT. The median age at the time of admission for AVM treatment was 14.6 (IQR 12-19, range 7-21 years) years. The median age at the time of the late follow-up DSA was 33.8 years (IQR 29.8-38.6, range 19.4-47.9 years). Two recurrent sporadic AVMs and one recurrent AVM in a patient with hereditary hemorrhagic telangiectasia (HHT) were detected. The recurrence rate was 4.9% for sporadic AVMs and 7.1% if HHT-AVM was included. All the recurrent AVMs had originally bled and been treated microsurgically. The patients with sporadic AVM recurrence had been smoking their whole adult lives. CONCLUSIONS: Pediatric and adolescent patients are prone to develop recurrent AVMs, even after complete AVM obliteration verified by angiography. Therefore, imaging follow-up is recommended.

Effect of Sleep Deprivation on Simulated Microsurgical Vascular Anastomosis.

BACKGROUND: The effects of sleep deprivation on surgical performance have been well documented. However, reports on the effects of sleep deprivation on microneurosurgery are limited. This study aimed to investigate the effects of sleep deprivation on microneurosurgery. METHODS: Ten neurosurgeons participated in the anastomosis of a vessel model using a microscope in sleep-deprived and normal states. We evaluated the procedure time (PT), stitch time (ST), interval time (IT), number of unachieved movements (NUM), leak rate, and practical scale for anastomosis quality assessment. Each parameter was compared between normal and sleep-deprived states. Subanalyses were performed on the two groups based on PT and NUM under the normal state (proficient and nonproficient groups). RESULTS: Although no significant differences were noted in PT, ST, NUM, leak rate, or practical scale, IT was significantly prolonged under sleep deprivation compared with the normal state (mean: 258.8 ± 94.0 vs. 199.3 ± 74.9 seconds; p = 0.02). IT was significantly prolonged under sleep deprivation in the nonproficient group based on both PT and NUM (PT: 234.2 ± 71.6 vs. 321.2 ± 44.7 seconds, p = 0.04; NUM: 173.3 ± 73.6 vs. 218.7 ± 97.7; p = 0.02), whereas no significant difference was observed in the proficient group (PT: 147.0 ± 47.0 vs. 165.3 ± 61.1 seconds, p = 0.25; NUM: 173.3 ± 73.6 vs. 218.7 ± 97.7; p = 0.25). CONCLUSIONS: Although IT was significantly prolonged under sleep deprivation in the nonproficient group, there was no decline in performance skills in either the proficient or nonproficient group. The effect of sleep deprivation may require caution in the nonproficient group, but it is possible that certain microneurosurgical outcomes can be achieved under sleep deprivation.

Treatment and Outcome of Anterior Inferior Cerebellar Artery (AICA) Aneurysms: Helsinki Series of 15 Consecutive Patients.

Objective Anterior inferior cerebellar artery (AICA) aneurysms are rare posterior circulation lesions that are challenging to treat. This article presents the treatment and clinical outcome of AICA aneurysms in an unselected cohort of patients. Methods A retrospective analysis of patient record files, digital subtraction angiography, and computed tomography angiography images of 15 consecutive patients harboring AICA aneurysms treated between 1968 and 2017. Results Of the 15 AICA aneurysm patients reviewed, 12 (80%) were females. Twenty percent had intracerebral hemorrhage and 40% presented with intraventricular hemorrhage. Eleven out of 15 (73%) patients presented with subarachnoid hemorrhage (SAH); 82% of SAH patients had a good-grade SAH (Hunt and Hess grade 1-3). Eleven patients (73%) were treated surgically, three (20%) were treated conservatively, and one (7%) had coil embolization. In 27% of patients, a subtemporal approach with anterior petrosectomy was performed. A retrosigmoid approach was used in the remaining 73%. In 18% of the patients, a parent vessel occlusion was necessary to occlude the aneurysm. Five out of 11 (47%) of the patients developed postoperative cranial nerve deficits. Twenty-seven percent developed shunt-dependent hydrocephalus. All patients who presented with an unruptured AICA aneurysm had good clinical outcome (modified Rankin scale [mRS] 1-2). In patients with SAH, 82% achieved good clinical outcome and 18% had poor clinical outcome (mRS 3-6) after 1 year. Conclusion Surgical treatment of AICA aneurysms has a high rate of cranial nerve deficits but most of patients have a good long-term clinical outcome.

3D Exoscopes in Experimental Microanastomosis: A Comparison of Different Systems.

Background: In recent years, three-dimensional exoscopes have been increasingly applied in neurosurgery. Multiple exoscopic systems are available, all offering specific features. In this study, we assessed practical and visualization performance of four different exoscopic systems in a highly challenging microsurgical procedure, and evaluated whether these affected the quality of work. Methods: We included four different exoscopes: Olympus ORBEYE, Zeiss KINEVO, Storz VITOM, and Aesculap AEOS. With each exoscope, ten experimental bypass procedures were carried out on chicken wing vessels at a depth of 3 cm. We evaluated the quality of the anastomoses, the practical considerations for the setup of the exoscopic systems, and the visualization quality by tracking the number of unnecessary movements. Results: All included exoscopes enabled us to perform the bypass procedures with mostly adequate or excellent results. Surgically, the most pronounced difference between the exoscopes was the duration of the procedure, which was mainly due to the number of unnecessary movements. Practically, the exoscopes differ highly which is important when considering which exoscope to apply. Conclusions: This is the first study comparing different exoscope systems while performing the same challenging microsurgical procedure. We found major practical differences between the exoscopes that determine the suitability of an exoscope based on the demands and conditions of the surgical procedure. Therefore, preprocedural practical training with the exoscope is required.

Small amount of alcohol did not deteriorate microsurgical dexterity: a prospective laboratory study.

BACKGROUND: Alcohol consumption has been reported to deteriorate surgical performance both immediately after consumption as well as on the next day. We studied the early effects of alcohol consumption on microsurgical manual dexterity in a laboratory setting. METHOD: Six neurosurgeons or neurosurgical residents (all male) performed micro- and macro suturing tasks after consuming variable amounts of alcohol. Each participant drank 0-4 doses of alcohol (14 g ethanol). After a delay of 60-157 min, he performed a macrosurgical and microsurgical task (with a surgical microscope). The tasks consisted of cutting and re-attaching a circular latex flap (diameter: 50 mm macrosuturing, 4 mm microsuturing) with eight interrupted sutures (4-0 multifilament macrosutures, 9-0 monofilament microsutures). We measured the time required to complete the sutures, and the amplitude and the frequency of physiological tremor during the suturing. In addition, we used a four-point ordinal scale to rank the quality of the sutures for each task. Each participant repeated the tasks several times on separate days varying the pre-task alcohol consumption (including one sober task at the end of the data collection). RESULTS: A total of 93 surgical tasks (47 macrosurgical, 46 microsurgical) were performed. The fastest microsurgical suturing (median 11 min 49 s, [interquartile range (IQR) 654 to 761 s]) was recorded after three doses of alcohol (median blood alcohol level 0.32‰). The slowest microsurgical suturing (median 15 min 19 s, [IQR 666 to 1121 s]) was observed after one dose (median blood alcohol level 0‰). The quality of sutures was the worst (mean 0.70 [standard deviation (SD) 0.48] quality points lost) after three doses of alcohol and the best (mean 0.33 [SD 0.52] quality points lost) after four doses (median blood alcohol level 0.44‰). CONCLUSIONS: Consuming small amount of alcohol did not deteriorate microsurgical performance in our study. An observed reduction in physiological tremor may partially explain this.

Correlation between arteriovenous malformation nidus size and intraparenchymal hematoma volume in the event of rupture.

•Our study did not support the hypothesis that small AVMs lead to larger hematoma volumes in the event of a rupture.•AVM size did not correlate with the clinical severity of the bleeding as measured with HH and WFNS scores.•Larger supratentorial hematomas were associated with a more severe clinical manifestation and a poorer outcome.•In the event of a rupture, the AVM size did not correlate with the 2-4-month GOS.•AVICH score showed a good correlation with the 2-4-month GOS score.

Predicting Therapeutic Response to Unfractionated Heparin Therapy: Machine Learning Approach.

BACKGROUND: Unfractionated heparin (UFH) is an anticoagulant drug that is considered a high-risk medication because an excessive dose can cause bleeding, whereas an insufficient dose can lead to a recurrent embolic event. Therapeutic response to the initiation of intravenous UFH is monitored using activated partial thromboplastin time (aPTT) as a measure of blood clotting time. Clinicians iteratively adjust the dose of UFH toward a target, indication-defined therapeutic aPTT range using nomograms, but this process can be imprecise and can take ≥36 hours to achieve the target range. Thus, a more efficient approach is required. OBJECTIVE: In this study, we aimed to develop and validate a machine learning (ML) algorithm to predict aPTT within 12 hours after a specified bolus and maintenance dose of UFH. METHODS: This was a retrospective cohort study of 3019 patient episodes of care from January 2017 to August 2020 using data collected from electronic health records of 5 hospitals in Queensland, Australia. Data from 4 hospitals were used to build and test ensemble models using cross-validation, whereas data from the fifth hospital were used for external validation. We built 2 ML models: a regression model to predict the aPTT value after a UFH bolus dose and a multiclass model to predict the aPTT, classified as subtherapeutic (aPTT <70 seconds), therapeutic (aPTT 70-100 seconds), or supratherapeutic (aPTT >100 seconds). Modeling was performed using Driverless AI (H2O), an automated ML tool, and 17 different experiments were iteratively conducted to optimize model accuracy. RESULTS: In predicting aPTT, the best performing model was an ensemble with 4x LightGBM models with a root mean square error of 31.35 (SD 1.37). In predicting the aPTT class using a repurposed data set, the best performing ensemble model achieved an accuracy of 0.599 (SD 0.0289) and an area under the receiver operating characteristic curve of 0.735. External validation yielded similar results: root mean square error of 30.52 (SD 1.29) for the aPTT prediction model, and accuracy of 0.568 (SD 0.0315) and area under the receiver operating characteristic curve of 0.724 for the aPTT multiclassification model. CONCLUSIONS: To the best of our knowledge, this is the first ML model applied to intravenous UFH dosing that has been developed and externally validated in a multisite adult general medical and surgical inpatient setting. We present the processes of data collection, preparation, and feature engineering for replication.

Fast Transition from Open Surgery to Endovascular Treatment of Unruptured Anterior Communicating Artery Aneurysms-A Retrospective Analysis of 128 Patients.

OBJECTIVE: Anterior communicating artery aneurysms (ACoAAs) are challenging to treat both surgically and endovascularly. In this study, we evaluate the treatment-related morbidity and clinical outcome of microsurgical clipping and endovascular treatment for a consecutive series of unruptured ACoAAs while the treatment paradigm was in transition from surgical to endovascular first. METHODS: We retrospectively reviewed clinical and radiologic data of adult patients who underwent microsurgical clipping or endovascular treatment of an unruptured ACoAA at a high-volume academic neurovascular center (Helsinki University Hospital) during 2012-2019. During this period, a transition from microsurgical clipping to endovascular treatment took place. Regarding outcome, we focused on treatment-related complications, discharge-to-home rates, functional performance (modified Rankin Scale score), and obliteration rates. RESULTS: Of 128 treated ACoAAs, 81 (64%) were treated surgically and 47 (36%) endovascularly. There was no difference in major complications, intracranial hemorrhagic complications or ischemic complications, discharge-to-home rates, or functional performance between the surgically and endovascularly treated patients. With time, a decrease in major complications was observed in the surgical cases (from 29% to 17%), whereas the major complication rate increased in the endovascularly patients (from 0% to 25%). Cerebral ischemia was the most frequent complication in both groups. The risk for permanent neurologic deficit remained low in both groups (9% for endovascular and 5% for surgery). CONCLUSIONS: We did not find any major differences regarding complications and outcomes after the treatment paradigm shift from clipping to endovascular of unruptured ACoAAs. Prospective studies evaluating durability of treatments are needed to compare overall effectiveness.

Explainable machine learning for real-time deterioration alert prediction to guide pre-emptive treatment.

The Electronic Medical Record (EMR) provides an opportunity to manage patient care efficiently and accurately. This includes clinical decision support tools for the timely identification of adverse events or acute illnesses preceded by deterioration. This paper presents a machine learning-driven tool developed using real-time EMR data for identifying patients at high risk of reaching critical conditions that may demand immediate interventions. This tool provides a pre-emptive solution that can help busy clinicians to prioritize their efforts while evaluating the individual patient risk of deterioration. The tool also provides visualized explanation of the main contributing factors to its decisions, which can guide the choice of intervention. When applied to a test cohort of 18,648 patient records, the tool achieved 100% sensitivity for prediction windows 2-8 h in advance for patients that were identified at 95%, 85% and 70% risk of deterioration.