Payer shifting after expansions in access to private care among veterans

OBJECTIVE

To investigate whether expanded access to Veterans Affairs (VA)-purchased care increased overall utilization or induced a shift from other payers to VA for emergency care among VA enrollees.

DATA SOURCES AND STUDY SETTING

This study included all emergency department (ED) encounters in 2019 from hospitals in the state of New York.

STUDY DESIGN

We conducted a difference-in-differences analysis comparing VA enrollees to the general population before and after the implementation of the Maintaining Internal Systems and Strengthening Integrated Outside Networks (MISSION) Act in June 2019.

DATA COLLECTION/EXTRACTION METHODS

We included all ED visits with individuals aged 30 or older at the time of the encounter. Individuals were considered eligible for the policy change if they were enrolled with VA at the beginning of 2019.

PRINCIPAL FINDINGS

Of the 5,577,199 ED visits in the sample, 4.9% (n = 253,799) were made by VA enrollees. Of these, 44.9% of visits were paid by Medicare, 32.8% occurred in VA facilities, and 7% were paid by private health insurance. There was a 6.4% (2.91 percentage points; std. error = 0.18; p < 0.01) decrease in the proportion of ED visits paid by Medicare among VA enrollees relative to the general population after the implementation of the MISSION Act in June 2019. This decrease was larger for ED visits with a subsequent inpatient admission (-8.4%; 4.87 percentage points; std. error = 0.33; p < 0.01). There was no statistically significant change in the total volume of ED visits (0.06%; std. error = 0.08; p = 0.45).

CONCLUSIONS

Leveraging a novel dataset, we demonstrate that MISSION Act implementation coincided with a shift in the financing of non-VA ED visits from Medicare to VA without any increase in overall ED utilization. These findings have important implications for VA health care financing and delivery.

Ensuring privacy protection in the era of big laparoscopic video data: development and validation of an inside outside discrimination algorithm (IODA)

BACKGROUND

Laparoscopic videos are increasingly being used for surgical artificial intelligence (AI) and big data analysis. The purpose of this study was to ensure data privacy in video recordings of laparoscopic surgery by censoring extraabdominal parts. An inside-outside-discrimination algorithm (IODA) was developed to ensure privacy protection while maximizing the remaining video data.

METHODS

IODAs neural network architecture was based on a pretrained AlexNet augmented with a long-short-term-memory. The data set for algorithm training and testing contained a total of 100 laparoscopic surgery videos of 23 different operations with a total video length of 207 h (124 min ± 100 min per video) resulting in 18,507,217 frames (185,965 ± 149,718 frames per video). Each video frame was tagged either as abdominal cavity, trocar, operation site, outside for cleaning, or translucent trocar. For algorithm testing, a stratified fivefold cross-validation was used.

RESULTS

The distribution of annotated classes were abdominal cavity 81.39%, trocar 1.39%, outside operation site 16.07%, outside for cleaning 1.08%, and translucent trocar 0.07%. Algorithm training on binary or all five classes showed similar excellent results for classifying outside frames with a mean F1-score of 0.96 ± 0.01 and 0.97 ± 0.01, sensitivity of 0.97 ± 0.02 and 0.0.97 ± 0.01, and a false positive rate of 0.99 ± 0.01 and 0.99 ± 0.01, respectively.

CONCLUSION

IODA is able to discriminate between inside and outside with a high certainty. In particular, only a few outside frames are misclassified as inside and therefore at risk for privacy breach. The anonymized videos can be used for multi-centric development of surgical AI, quality management or educational purposes. In contrast to expensive commercial solutions, IODA is made open source and can be improved by the scientific community.

680. Utility of Cefoxitin for the Treatment of ESBL-producing E. coli Urinary Tract Infections

Background

Carbapenems are regarded as first-line agents for the treatment of ESBL-producing bacterial infections, even when in vitro activity to other beta lactams is demonstrated. However, overuse is associated with the selection of beta-lactamases that can hydrolyze carbapenems. Patients with isolated urinary tract infections due to ESBL–producing E. coli potentially have a lower burden of infection ("low inoculum"). These patients are frequently less ill and may benefit from non-carbapenem therapy. Cephamycins have in vitro activity against ESBL–producing Enterobacterales, and may be useful in the treatment of ESBL E. coli urinary tract infections as a carbapenem sparing agent.

Methods

Patients 18 years of age and older, hemodynamically stable with negative blood cultures and who demonstrated signs and symptoms of a urinary tract infection with a urine culture showing ESBL E. coli, were eligible to participate. Patients were started on initial antimicrobial therapy at the discretion of the admitting physician, and if the urine culture showed ESBL E. coli with an MIC of less than 8, the antimicrobial stewardship team requested to switch the patient to Cefoxitin 2 g intravenously every 6 hours, via extended infusion. The dose was renally adjusted as appropriate, and patients with beta-lactam allergies were excluded.

Results

46 patients were evaluated between May 2017-April 2022. Patients received antibiotics for 2.6+1 day before the switch to Cefoxitin. The mean duration of therapy with Cefoxitin was 4.5+2.2 days. All 46 patients had resolution of fever and dysuria. Six patients were readmitted within 30 days, but none were due to a urinary tract infection. 30 patients had a repeat urinalysis at 48 hours, and all demonstrated reduction in pyuria. 39 patients had a repeat urine culture following completion of therapy, and 36 had converted to culture-negative status.

Conclusion

Cefoxitin is a useful agent for the treatment of ESBL E. coli urinary tract infections. The optimal dose appears to be 2 g intravenously every 6 hours, via extended infusion, in patients with normal creatinine clearance. Further larger studies, involving patients with pyelonephritis, are needed to validate these findings.

Disclosures

Ramesh V. Nathan, MD, Eli Lilly: Grant/Research Support|PPD: Grant/Research Support.

Diagnostic performance of MRI perfusion and spectroscopy for brainstem glioma grading

OBJECTIVE

This study investigated the roles of dynamic susceptibility contrast (DSC) perfusion and multivoxel magnetic resonance spectroscopy (MRS) in grading brainstem glioma (BSG).

PATIENTS AND METHODS

Our retrospective study comprised 12 patients, including 6 with pathology verified low-grade BSGs and 6 with high-grade BSGs. We examined differences in age, relative cerebral blood volume (rCBV), regional cerebral blood flow (rCBF), and the metabolite ratios of choline (Cho)/N-acetyl aspartate (NAA) and Cho/creatine (Cr) between these two groups using the Mann-Whitney U test and Chi-square test. Receiver operating characteristic (ROC) curve analysis was used to establish cutoff values and assess their usefulness in grading BSG.

RESULTS

The Cho/NAA metabolite ratio had the strongest preoperative predictive performance for identifying the correct histological grade among BSGs, with an area under the ROC curve (AUC) value of 0.944 (cutoff: 3.88, sensitivity [Se]: 83.3%; specificity [Sp]: 100%), followed by the Cho/Cr ratio (cutoff: 3.08; AUC: 0.917; Se: 83.3%; Sp: 100%), rCBF (cutoff: 3.56, AUC: 0.917; Se: 83.3%; Sp: 100%), rCBV (cutoff: 3.16, AUC: 0.889; Se: 100%; Sp: 66.7%), and age (cutoff: 9.5 years, AUC: 0.889; Se: 100%; Sp: 83.3%).

CONCLUSIONS

rCBF and rCBV values comparing solid tumors with the normal brain parenchyma and the metabolite ratios for Cho/NAA and Cho/Cre may serve as useful indices for establishing BSG grading and provide important information when determining treatment planning and prognosis in patients with BSG.

Processed electroencephalography: impact of patient age and surgical position on intraoperative processed electroencephalogram monitoring of burst-suppression

We previously reported that processed EEG underestimated the amount of burst suppression compared to off-line visual analysis. We performed a follow-up study to evaluate the reasons for the discordance. Forty-five patients were monitored intraoperatively with processed EEG. A computer algorithm was used to convert the SedLine® (machine)-generated burst suppression ratio into a raw duration of burst suppression. The reference standard was a precise off-line measurement by two neurologists. We measured other potential variables that may affect machine accuracy such as age, surgery position, and EEG artifacts. Overall, the median duration of bust suppression for all study subjects was 15.4 min (Inter-quartile Range [IQR] = 1.0-20.1) for the machine vs. 16.1 min (IQR = 0.3-19.7) for the neurologists' assessment; the 95% limits of agreement fall within - 4.86 to 5.04 s for individual 30-s epochs. EEG artifacts did not affect the concordance between the two methods. For patients in prone surgical position, the machine estimates had significantly lower overall sensitivity (0.86 vs. 0.97; p = 0.038) and significantly wider limits of agreement ([- 4.24, 3.82] seconds vs. [- 1.36, 1.13] seconds, p = 0.001) than patients in supine position. Machine readings for younger patients (age < 65 years) had higher sensitivity (0.96 vs 0.92; p = 0.021) and specificity (0.99 vs 0.88; p = 0.007) for older patients. The duration of burst suppression estimated by the machine generally had good agreement compared with neurologists' estimation using a more precise off-line measurement. Factors that affected the concordance included patient age and position during surgery, but not EEG artifacts.

Randomized comparison between perineural dexamethasone and combined perineural dexamethasone-dexmedetomidine for ultrasound-guided infraclavicular block

BACKGROUND

This randomized trial compared perineural dexamethasone with combined perineural dexamethasone-dexmedetomidine for ultrasound-guided infraclavicular block. We hypothesized that the combination of perineural adjuvants would result in a longer motor block.

METHODS

Fifty patients undergoing upper limb surgery with ultrasound-guided infraclavicular block (using 35 mL of lidocaine 1%-bupivacaine 0.25% with epinephrine 5 µg/mL) were randomly allocated to receive perineural dexamethasone (2 mg) or combined perineural dexamethasone (2 mg)-dexmedetomidine (50 µg). After the performance of the block, a blinded observer assessed the success rate (defined as a minimal sensorimotor composite score of 14 out of 16 points at 30 min), the onset time (defined as the time required to reach a minimal composite score of 14 points) as well as the incidence of surgical anesthesia (defined as the ability to complete surgery without local infiltration, supplemental blocks, intravenous opioids or general anesthesia).Postoperatively, the blinded observer contacted patients with successful blocks to inquire about the duration of motor block, sensory block and postoperative analgesia.

RESULTS

No intergroup differences were observed in terms of success rate, onset time and surgical anesthesia. Compared with dexamethasone alone, combined dexamethasone-dexmedetomidine provided longer durations of motor block (21.5 (2.7) vs 17.0 (3.9) hours; p<0.001; 95% CI 2.6 to 6.4), sensory block (21.6 (3.6) vs 17.2 (3.6) hours; p<0.001; 95% CI 2.2 to 6.5), and postoperative analgesia (25.5 (9.4) vs 23.5 (5.6) hours; p=0.038; 95% CI 1.0 to 7.7).

CONCLUSION

Compared with perineural dexamethasone (2 mg) alone, combined perineural dexamethasone (2 mg)-dexmedetomidine (50 µg) results in longer durations of sensorimotor block and analgesia. Further studies are required to determine the optimal dosing combination for dexamethasone-dexmedetomidine.

TRIAL REGISTRATION NUMBER

ClinicalTrials.gov identifier: NCT04875039.

POS1405 LESS THAN 50% FEMALES WITH CHRONIC RHEUMATIC INFLAMMATORY DISEASES CONTINUE A DMARD DURING PREGNANCY: A DESCRIPTIVE ANALYSIS OF THE NATIONAL FRENCH SOCIAL SECURITY DATABASE

Background

Treatment of patients with chronic rheumatic inflammatory diseases (CRID) during pregnancy has changed in the last decade, namely due to the availability of new DMARDs labelled to be used during pregnancy.

Objectives

To describe the anti-rheumatic drug use during pregnancy in women with CRID (i.e. rheumatoid arthritis (RA) or spondyloarthritis (SpA)) in France over the past decade.

Methods

This is a retrospective cohort study within the French Healthcare database (SNDS), which covers 98% of the French population. Adult women were included if they had RA or SpA according to CIM-10 codes, had started a singleton pregnancy between 2008 and 2017 (index date), and were continuously covered by this health insurance from 1-year before pregnancy onset to 1-year after end of the pregnancy or death (whichever comes first). The treatment exposures of interest were: NSAIDs, oral corticosteroids, csDMARD (methotrexate, leflunomide, sulfasalazine, azathioprine, hydroxychloroquine), biologics (anti-TNF, rituximab, abatacept, tocilizumab, ustekinumab, anakinra). Exposure during pregnancy was defined as at least one drug reimbursement from the 6 months before the last menstrual period (LMP) to the end of pregnancy period.

Results

Among the 35,737 adult women with a CIRD (40.7% with RA and 59.3% with SpA) who had a past history of DMARD reimbursement, 11,274 (41.7%) started a singleton pregnancy during the study period. In total, during preconception and pregnancy, 4,773 (42.3%) women were not delivered any DMARD nor corticosteroids, 769 (6.8%) were delivered corticosteroids alone, 3,639 (32.2%) a csDMARD alone and 2,862 (25.4%) a biologic (among whom 33.1% associated a csDMARD). Biologics delivered during pregnancy were mainly anti-TNFs (92.1%).

Exposure to NSAIDs was more frequent during the first trimester (30% patients) of pregnancy but occurred all along the pregnancy (6% and 2% in the second and third trimesters, respectively). Conversely, exposure to oral corticosteroids was stable during the pregnancy (33% to 27%); however, more than half of the prescriptions corresponded to doses higher than 10mg. Exposure to DMARDs including bDMARDs during pregnancy was more frequent during the first trimester, compared to the rest of the pregnancy (see graph).

Conclusion

Overall, less than 50% of women with a CRID who received a DMARD prior to the pregnancy continued to retrieve such treatment during pregnancy, and overall less than 20% were delivered biologics during pregnancy. Whether the withdrawal of DMARDs led to unfavorable maternal and pregnancy outcomes needs to be evaluated.

Acknowledgements

This study was conducted thanks to a grant from the French Ministry of Health - Programme Hospitalier de Recherche Clinique

Disclosure of Interests

Anna Moltó Consultant of: UCB, Abbvie, Lilly, Pfizer, BMS, MSD, Novartis, Biogen, Janssen, Grant/research support from: UCB, Aya Ajrouche: None declared, Diep Tran: None declared, Barbara Roux: None declared, Nathalie Costedoat-Chalumeau Grant/research support from: UCB, Elisabeth Elefant: None declared, Vassilis Tsatsaris: None declared, Jeanne Fresson: None declared, Brigitte Bader-Meunier: None declared, Bruno Fautrel: None declared, Florence Tubach: None declared

AB1391 19% PATIENTS WITH CHRONIC RHEUMATIC INFLAMMATORY DISEASES PRESENT AN UNFAVORABLE PREGNANCY OUTCOME: A DESCRIPTIVE ANALYSIS OF THE NATIONAL FRENCH SOCIAL SECURITY DATABASE

Background

patients with chronic rheumatic inflammatory diseases (CRID, i.e. rheumatoid arthritis (RA) or spondyloarthritis (SpA)) have been reported to have poorer pregnancy outcomes than the general population.

Objectives

to describe the pregnancy outcomes of singleton in patients with CRID in France in the past decade.

Methods

this is an analysis of the French Health Insurance claims database (SNDS), which represents 87% of the French population) from 2008 to 2016. To be included in the analysis patients had to be identified as RA or SpA according to existing diagnostic algorithms, to have at least one pregnancy declaration in the database, and to have continuous health insurance from 1-year before pregnancy onset to 1-year after end of the pregnancy or until death (if death occured before the 1-year-period after the end of pregnancy) to be included. Only the first singleton pregnancy occurring during the study period was included in this analysis. Both maternal and pregnancy outcomes were considered. Outcomes were identified either by ICM-10 codes or hospital discharge summaries discharge between 2008-2016.

Results

Among the 35 737 identified adult females with a CRID diagnosis (40.7% with RA and 59.3% with SpA), 27 722 (78%) had a pregnancy during the study period. 11 274 (42%) had received at least one DMARD prior to the pregnancy and were included in the analysis. Among them, only 4025 (36%) were exposed to DMARDs during pregnancy.

Mean (SD) age of females at the start of the pregnancy was 32 (5) years, and mean (SD) disease duration was 4 (4) years. Pregnancy ended before 13 WG in 21% and after 37 WG in 70% cases. Live-birth represented the most frequent pregnancy outcome (76.9%), and overall 34.7% patients presented at least one unfavorable outcome (see Table 1).

Table 1.Unfavourable outcomeN(%)Pregnancy outcomesMiscarriage (<22WG)579 (5.3%)Abortion91 (0.8%)Stillbirth48 (0.4%)Preterm delivery (>= 22WG and <37 WG)779 (7.1%)Low weight at birth (<3th percentile)287 (2.6%)Perinatal mortality (22WG to 6 days of life)2 (0%)Maternal outcomesSevere maternal infection*211 (1.9%)Newborn outcomesHospitalisation in neonatal intensive care > 48h in full-term newborns (after 37 WG)95 (0.8%)Neonatal mortality (27 first days of life)2 (0%)Severe infection * during first year603 (5.3%)

Conclusion

More than 75% pregnancies in patients with CRID resulted in a live-birth. Prevalence of miscarriage was surprisingly low, probably related to under-coding. Maternal and infant outcomes seemed comparable to general population. Whether medications had an impact on such outcomes is currently under evaluation.

Disclosure of Interests

None declared

Real-world Safety of Bevacizumab with First-line Combination Chemotherapy in Patients with Metastatic Colorectal Cancer: Population-based Retrospective Cohort Studies in Three Canadian Provinces

AIMS

To examine the real-world safety of adding bevacizumab to first-line irinotecan-based chemotherapy for patients with metastatic colorectal cancer (mCRC).

MATERIALS AND METHODS

Patients diagnosed with CRC in three Canadian provinces (Ontario, Saskatchewan and British Columbia) who received publicly funded bevacizumab and/or irinotecan from 2000 to 2016 were identified from cancer registries. Propensity score 1:1 matching (PSM) and inverse probability of treatment weighting (IPTW) were performed to contemporaneous and historical controls, adjusting for baseline demographic and clinical characteristics. Safety end points evaluated during first-line treatment plus 30 days included mortality within 30 days and all-cause-, chemotherapy- and bevacizumab-related hospitalisations. Chemotherapy- and bevacizumab-related visits were defined as hospitalisations for specific conditions commonly associated with chemotherapy (e.g. infections) or bevacizumab (e.g. arteriovenous thromboembolism) using most responsible diagnosis codes. In PSM and IPTW-weighted cohorts, we assessed event frequencies using odds ratios from logistic regressions and event rate ratios using negative binomial regression models. The results from each province and comparison were pooled using random-effects meta-analysis.

RESULTS

We identified 16 250 mCRC patients who received first-line irinotecan-based treatment. In PSM cohorts, bevacizumab was associated with fewer deaths within 30 days of treatment compared with contemporaneous (pooled odds ratio = 0.62; 95% confidence interval 0.50-0.75) and historical controls (pooled odds ratio = 0.73; 95% confidence interval 0.58-0.93). Hospitalisations were more frequent among patients treated with bevacizumab compared with historical controls but similar to contemporaneous controls. As patients receiving bevacizumab were exposed to a longer average treatment duration, across their full treatment duration, patients receiving bevacizumab had significantly lower rates of hospitalisations (contemporaneous pooled rate ratio = 0.56; 95% confidence interval 0.47-0.67; historical pooled rate ratio = 0.73; 95% confidence interval 0.56-0.95). Similar trends were observed for chemotherapy- and bevacizumab-related hospitalisations and in IPTW-weighted cohorts.

DISCUSSION

We did not observe any increase in rates of hospitalisation or death within 30 days of treatment among mCRC patients treated with bevacizumab plus chemotherapy versus chemotherapy alone; these findings should be interpreted with caution due to the risk of residual confounding.

Preoperative left ventricular end-systolic dimension predicts occurrence of aortic insufficiency following aortic valve preservation and repair surgery

BACKGROUND

Preoperative left ventricular (LV) end-systolic dimension (LVESD) ≥5.0 cm is a class IIa indication for surgical intervention for aortic insufficiency (AI); however, the effect of LV dilatation on the longevity of the aortic valve (AV) has not yet been investigated. This study aimed to assess the impact of preoperative LV dimension on the long-term outcome of AV preservation surgery.

METHODS

Between 2009 and 2019, 256 patients underwent AV preservation surgery at a single center. The median duration of follow-up was 5 years. The primary outcome was the development of >1+ AI at 6 years; secondary outcomes include long-term mortality, freedom from >2+ AI, and freedom from AV reoperation. Cox proportional hazard analysis was performed to identify predictors of AV deterioration.

RESULTS

In-hospital mortality was 0.8%, and mean survival at 8 years was 85.5 ± 3.4%. Mean freedom from >1+ AI at 6 years was 71.1 ± 3.4%. Patients with preoperative indexed LVESD (LVESDi) ≥2.0 cm/m² were at greater risk of developing >1+ AI at 6 years compared with patients with preoperative LVESDi of 1.5 to 1.9 cm/m² and ≤1.4 cm/m² (50.3 ± 0.1% vs 80.9 ± 0.1% vs 92.2 ± 0.1%, respectively; P < .01). On risk-adjusted multivariable analysis, preoperative LVESDi was an independent predictor for recurrence of >1+ AI (hazard ratio, 2.2; 95% confidence interval, 1.5-3.4).

CONCLUSIONS

Preoperative LVESDi ≥2 cm/m² is associated with increased risk of recurrent >1+ AI following AV preservation surgery. Further investigation of the appropriate operative threshold for AI may be warranted.

Effect of Fibrinogen Concentrate vs Cryoprecipitate on Blood Component Transfusion After Cardiac Surgery: The FIBRES Randomized Clinical Trial

IMPORTANCE

Excessive bleeding is a common complication of cardiac surgery. An important cause of bleeding is acquired hypofibrinogenemia (fibrinogen level <1.5-2.0 g/L), for which guidelines recommend fibrinogen replacement with cryoprecipitate or fibrinogen concentrate. The 2 products have important differences, but comparative clinical data are lacking.

OBJECTIVE

To determine if fibrinogen concentrate is noninferior to cryoprecipitate for treatment of bleeding related to hypofibrinogenemia after cardiac surgery.

DESIGN, SETTING, AND PARTICIPANTS

Randomized clinical trial at 11 Canadian hospitals enrolling adult patients experiencing clinically significant bleeding and hypofibrinogenemia after cardiac surgery (from February 10, 2017, to November 1, 2018). Final 28-day follow-up visit was completed on November 28, 2018.

INTERVENTIONS

Fibrinogen concentrate (4 g; n = 415) or cryoprecipitate (10 units; n = 412) for each ordered dose within 24 hours after cardiopulmonary bypass.

MAIN OUTCOMES AND MEASURES

Primary outcome was blood components (red blood cells, platelets, plasma) administered during 24 hours post bypass. A 2-sample, 1-sided test for the ratio of the mean number of units was conducted to evaluate noninferiority (threshold for noninferiority ratio, <1.2).

RESULTS

Of 827 randomized patients, 735 (372 fibrinogen concentrate, 363 cryoprecipitate) were treated and included in the primary analysis (median age, 64 [interquartile range, 53-72] years; 30% women; 72% underwent complex operations; 95% moderate to severe bleeding; and pretreatment fibrinogen level, 1.6 [interquartile range, 1.3-1.9] g/L). The trial met the a priori stopping criterion for noninferiority at the interim analysis after 827 of planned 1200 patients were randomized. Mean 24-hour postbypass allogeneic transfusions were 16.3 (95% CI, 14.9 to 17.8) units in the fibrinogen concentrate group and 17.0 (95% CI, 15.6 to 18.6) units in the cryoprecipitate group (ratio, 0.96 [1-sided 97.5% CI, -∞ to 1.09; P < .001 for noninferiority] [2-sided 95% CI, 0.84 to 1.09; P = .50 for superiority]). Thromboembolic events occurred in 26 patients (7.0%) in the fibrinogen concentrate group and 35 patients (9.6%) in the cryoprecipitate group.

CONCLUSIONS AND RELEVANCE

In patients undergoing cardiac surgery who develop clinically significant bleeding and hypofibrinogenemia after cardiopulmonary bypass, fibrinogen concentrate is noninferior to cryoprecipitate with regard to number of blood components transfused in a 24-hour period post bypass. Use of fibrinogen concentrate may be considered for management of bleeding in patients with acquired hypofibrinogenemia in cardiac surgery.

TRIAL REGISTRATION

ClinicalTrials.gov Identifier: NCT03037424.

Deep learning as a predictive tool for fetal heart pregnancy following time-lapse incubation and blastocyst transfer

STUDY QUESTION

Can a deep learning model predict the probability of pregnancy with fetal heart (FH) from time-lapse videos?

SUMMARY ANSWER

We created a deep learning model named IVY, which was an objective and fully automated system that predicts the probability of FH pregnancy directly from raw time-lapse videos without the need for any manual morphokinetic annotation or blastocyst morphology assessment.

WHAT IS KNOWN ALREADY

The contribution of time-lapse imaging in effective embryo selection is promising. Existing algorithms for the analysis of time-lapse imaging are based on morphology and morphokinetic parameters that require subjective human annotation and thus have intrinsic inter-reader and intra-reader variability. Deep learning offers promise for the automation and standardization of embryo selection.

STUDY DESIGN, SIZE, DURATION

A retrospective analysis of time-lapse videos and clinical outcomes of 10 638 embryos from eight different IVF clinics, across four different countries, between January 2014 and December 2018.

PARTICIPANTS/MATERIALS, SETTING, METHODS

The deep learning model was trained using time-lapse videos with known FH pregnancy outcome to perform a binary classification task of predicting the probability of pregnancy with FH given time-lapse video sequence. The predictive power of the model was measured using the average area under the curve (AUC) of the receiver operating characteristic curve over 5-fold stratified cross-validation.

MAIN RESULTS AND THE ROLE OF CHANCE

The deep learning model was able to predict FH pregnancy from time-lapse videos with an AUC of 0.93 [95% CI 0.92-0.94] in 5-fold stratified cross-validation. A hold-out validation test across eight laboratories showed that the AUC was reproducible, ranging from 0.95 to 0.90 across different laboratories with different culture and laboratory processes.

LIMITATIONS, REASONS FOR CAUTION

This study is a retrospective analysis demonstrating that the deep learning model has a high level of predictability of the likelihood that an embryo will implant. The clinical impacts of these findings are still uncertain. Further studies, including prospective randomized controlled trials, are required to evaluate the clinical significance of this deep learning model. The time-lapse videos collected for training and validation are Day 5 embryos; hence, additional adjustment would need to be made for the model to be used in the context of Day 3 transfer.

WIDER IMPLICATIONS OF THE FINDINGS

The high predictive value for embryo implantation obtained by the deep learning model may improve the effectiveness of previous approaches used for time-lapse imaging in embryo selection. This may improve the prioritization of the most viable embryo for a single embryo transfer. The deep learning model may also prove to be useful in providing the optimal order for subsequent transfers of cryopreserved embryos.

STUDY FUNDING/COMPETING INTEREST(S)

D.T. is the co-owner of Harrison AI that has patented this methodology in association with Virtus Health. P.I. is a shareholder in Virtus Health. S.C., P.I. and D.G. are all either employees or contracted with Virtus Health. D.G. has received grant support from Vitrolife, the manufacturer of the Embryoscope time-lapse imaging used in this study. The equipment and time for this study have been jointly provided by Harrison AI and Virtus Health.

The association between platelet dysfunction and adverse outcomes in cardiac surgical patients

Haemostatic activation during cardiopulmonary bypass is associated with prothrombotic complications. Although it is not possible to detect and quantify haemostatic activation directly, platelet dysfunction, as measured with point-of-care-assays, may be a useful surrogate. In this study, we assessed the association between cardiopulmonary bypass-associated platelet dysfunction and adverse outcomes in 3010 cardiac surgical patients. Platelet dysfunction, as measured near the end of the rewarming phase of cardiopulmonary bypass, was calculated as the proportion of non-functional platelets after activation with collagen. Logistic regression and multivariable analyses were applied to assess the relationship between platelet dysfunction and a composite of in-hospital death; myocardial infarction; stroke; deep vein thrombosis or pulmonary embolism; and acute kidney injury (greater than a two-fold increase in creatinine). The outcome occurred in 251 (8%) of 3010 patients. The median (IQR [range]) percentage platelet dysfunction was less for those without the outcome as compared with those with the outcome; 14% (8-28% [1-99%]) vs. 19% (11-45% [2-98%]), p < 0.001. After risk adjustment, platelet dysfunction was independently associated with the composite outcome (p < 0.001), such that for each 1% increase in platelet dysfunction there was an approximately 1% increase in the composite outcome (OR 1.012; 95%CI 1.006-1.018). This exploratory study suggests that cardiopulmonary bypass-associated platelet dysfunction has prognostic value and may be a useful clinical measure of haemostatic activation in cardiac surgery.

Bioactive extracellular compounds produced by the dinoflagellate Alexandrium minutum are highly detrimental for oysters

Blooms of the dinoflagellate Alexandrium spp., known as producers of paralytic shellfish toxins (PSTs), are regularly detected on the French coastline. PSTs accumulate into harvested shellfish species, such as the Pacific oyster Crassostrea gigas, and can cause strong disorders to consumers at high doses. The impacts of Alexandrium minutum on C. gigas have often been attributed to its production of PSTs without testing separately the effects of the bioactive extracellular compounds (BECs) with allelopathic, hemolytic, cytotoxic or ichthyotoxic properties, which can also be produced by these algae. The BECs, still uncharacterized, are excreted within the environment thereby impacting not only phytoplankton, zooplankton but also marine invertebrates and fishes, without implicating any PST. The aim of this work was to compare the effects of three strains of A. minutum producing either only PSTs, only BECs, or both PSTs and BECs, on the oyster C. gigas. Behavioral and physiological responses of oysters exposed during 4 days were monitored and showed contrasted behavioral and physiological responses in oysters supposedly depending on produced bioactive substances. The non-PST extracellular-compound-producing strain primarily strongly modified valve-activity behavior of C. gigas and induced hemocyte mobilization within the gills, whereas the PST-producing strain caused inflammatory responses within the digestive gland and disrupted the daily biological rhythm of valve activity behavior. BECs may therefore have a significant harmful effect on the gills, which is one of the first organ in contact with the extracellular substances released in the water by A. minutum. Conversely, the PSTs impact the digestive gland, where they are released and mainly accumulated, after degradation of algal cells during digestion process of bivalves. This study provides a better understanding of the toxicity of A. minutum on oyster and highlights the significant role of BECs in this toxicity calling for further chemical characterization of these substances.