RESUMO
OBJECTIVE: Survival analysis is widely utilized in healthcare to predict the timing of disease onset. Traditional methods of survival analysis are usually based on Cox Proportional Hazards model and assume proportional risk for all subjects. However, this assumption is rarely true for most diseases, as the underlying factors have complex, non-linear, and time-varying relationships. This concern is especially relevant for pregnancy, where the risk for pregnancy-related complications, such as preeclampsia, varies across gestation. Recently, deep learning survival models have shown promise in addressing the limitations of classical models, as the novel models allow for non-proportional risk handling, capturing nonlinear relationships, and navigating complex temporal dynamics. METHODS: We present a methodology to model the temporal risk of preeclampsia during pregnancy and investigate the associated clinical risk factors. We utilized a retrospective dataset including 66,425 pregnant individuals who delivered in two tertiary care centers from 2015 to 2023. We modeled the preeclampsia risk by modifying DeepHit, a deep survival model, which leverages neural network architecture to capture time-varying relationships between covariates in pregnancy. We applied time series k-means clustering to DeepHit's normalized output and investigated interpretability using Shapley values. RESULTS: We demonstrate that DeepHit can effectively handle high-dimensional data and evolving risk hazards over time with performance similar to the Cox Proportional Hazards model, achieving an area under the curve (AUC) of 0.78 for both models. The deep survival model outperformed traditional methodology by identifying time-varied risk trajectories for preeclampsia, providing insights for early and individualized intervention. K-means clustering resulted in patients delineating into low-risk, early-onset, and late-onset preeclampsia groups-notably, each of those has distinct risk factors. CONCLUSION: This work demonstrates a novel application of deep survival analysis in time-varying prediction of preeclampsia risk. Our results highlight the advantage of deep survival models compared to Cox Proportional Hazards models in providing personalized risk trajectory and demonstrating the potential of deep survival models to generate interpretable and meaningful clinical applications in medicine.
Assuntos
Pré-Eclâmpsia , Humanos , Pré-Eclâmpsia/mortalidade , Gravidez , Feminino , Análise de Sobrevida , Fatores de Risco , Aprendizado Profundo , Adulto , Estudos Retrospectivos , Modelos de Riscos Proporcionais , Redes Neurais de Computação , Medição de Risco/métodosRESUMO
OBJECTIVE: Pain is a variably experienced symptom during pregnancy, and women scheduled for cesarean delivery, an increasingly common procedure, are a relatively understudied group who might be at higher pain risk. Although biopsychosocial factors are known to modulate many types of chronic pain, their contribution to late pregnancy pain has not been comprehensively studied. We aimed to identify biopsychosocial factors associated with greater pain severity and interference during the last week of pregnancy. METHODS: In this prospective, observational study, 662 pregnant women scheduled for cesarean delivery provided demographic and clinical information and completed validated psychological and pain assessments. Multivariable hierarchical linear regressions assessed independent associations of demographic, clinical, and psychological characteristics with pain severity and pain interference during the last week of pregnancy. RESULTS: Women in the study had a mean age of 34 years, and 73% identified as White, 11% as African American, 10% as Hispanic/Latina, and 6% as Asian. Most women (66%) were scheduled for repeat cesarean delivery. Significant independent predictors of worse pain outcomes included identifying as African American or Hispanic/Latina and having greater depression, sleep disturbance, and pain catastrophizing. Exploratory analyses showed that women scheduled for primary (versus repeat) cesarean delivery reported higher levels of anxiety and pain catastrophizing. CONCLUSIONS: Independent of demographic or clinical factors, psychological factors, including depression, sleep disturbance, and pain catastrophizing, conferred a greater risk of late pregnancy pain. These findings suggest that women at higher risk of pain during late pregnancy could benefit from earlier nonpharmacological interventions that concurrently focus on psychological and pain symptoms.
Assuntos
Cesárea , Dor Crônica , Gravidez , Feminino , Humanos , Adulto , Medição da Dor , Estudos Prospectivos , Catastrofização/psicologiaRESUMO
Preoperative anxiety has an incidence of 11-80% in patients undergoing surgical or interventional procedures. Understanding the role of preoperative anxiety on intraoperative anesthetic requirements and postoperative analgesic consumption would allow personalized anesthesia care. Over- or under-anesthetizing patients can lead to complications such as postoperative cognitive dysfunction in elderly patients, or procedural discomfort, respectively. Our scoping review focuses on the current evidence regarding the association between preoperative anxiety and intraoperative anesthetic and/or postoperative analgesic consumption in patients undergoing elective surgical or interventional procedures. Based on 44 studies that met the inclusion criteria, we found that preoperative anxiety has a significant positive correlation effect on intraoperative propofol and postoperative opioid consumption. The analysis of the literature is limited by the heterogeneity of preoperative anxiety tools used, study designs, data analyses, and outcomes. The use of shorter, validated preoperative anxiety assessment tools may help optimize the intraoperative anesthetic and postoperative analgesic regimen. Further research to determine the most feasible and clinically relevant preoperative anxiety tool and subsequent implementation has the potential to optimize perioperative care and improve patient outcomes.
Assuntos
Anestésicos , Propofol , Idoso , Analgésicos , Ansiedade/tratamento farmacológico , Procedimentos Cirúrgicos Eletivos , Humanos , Dor Pós-Operatória/tratamento farmacológicoRESUMO
PURPOSE OF REVIEW: Recent advancements in big data analytical tools and large patient databases have expanded tremendously the opportunities to track patient and safety outcomes.We discuss the strengths and limitations of large databases and implementation in practice with a focus on the current opportunities to use technological advancements to improve patient safety. RECENT FINDINGS: The most used sources of data for large patient safety observational studies are administrative databases, clinical registries, and electronic health records. These data sources have enabled research on patient safety topics ranging from rare adverse outcomes to large cohort studies of the modalities for pain control and safety of medications. Implementing the insights from big perioperative data research is augmented by automating data collection and tracking the safety outcomes on a provider, institutional, national, and global level. In the near future, big data from wearable devices, physiological waveforms, and genomics may lead to the development of personalized outcome measures. SUMMARY: Patient safety research using large databases can provide actionable insights to improve outcomes in the perioperative setting. As datasets and methods to gain insights from those continue to grow, adopting novel technologies to implement personalized quality assurance initiatives can significantly improve patient care.
Assuntos
Big Data , Registros Eletrônicos de Saúde , Humanos , Bases de Dados Factuais , Sistema de Registros , Avaliação de Resultados em Cuidados de Saúde/métodosRESUMO
BACKGROUND: Detailed reviews of closed malpractice claims have provided insights into the most common events resulting in litigation and helped improve anesthesia care. In the past 10 years, there have been multiple safety advancements in the practice of obstetric anesthesia. We investigated the relationship among contributing factors, patient injuries, and legal outcome by analyzing a contemporary cohort of closed malpractice claims where obstetric anesthesiology was the principal defendant. METHODS: The Controlled Risk Insurance Company (CRICO) is the captive medical liability insurer of the Harvard Medical Institutions that, in collaboration with other insurance companies and health care entities, contributes to the Comparative Benchmark System database for research purposes. We reviewed all (N = 106) closed malpractice cases related to obstetric anesthesia between 2005 and 2015 and compared the following classes of injury: maternal death and brain injury, neonatal death and brain injury, maternal nerve injury, and maternal major and minor injury. In addition, settled claims were compared to the cases that did not receive payment. χ, analysis of variance, Student t test, and Kruskal-Wallis tests were used for comparison between the different classes of injury. RESULTS: The largest number of claims, 54.7%, involved maternal nerve injury; 77.6% of these claims did not receive any indemnity payment. Cases involving maternal death or brain injury comprised 15.1% of all cases and were more likely to receive payment, especially in the high range (P = .02). The most common causes of maternal death or brain injury were high neuraxial blocks, embolic events, and failed intubation. Claims for maternal major and minor injury were least likely to receive payment (P = .02) and were most commonly (34.8%) associated with only emotional injury. Compared to the dropped/denied/dismissed claims, settled claims more frequently involved general anesthesia (P = .03), were associated with delays in care (P = .005), and took longer to resolve (3.2 vs 1.3 years; P < .0001). CONCLUSIONS: Obstetric anesthesia remains an area of significant malpractice liability. Opportunities for practice improvement in the area of severe maternal injury include timely recognition of high neuraxial block, availability of adequate resuscitative resources, and the use of advanced airway management techniques. Anesthesiologists should avoid delays in maternal care, establish clear communication, and follow their institutional policy regarding neonatal resuscitation. Prevention of maternal neurological injury should be directed toward performing neuraxial techniques at the lowest lumbar spine level possible and prevention/recognition of retained neuraxial devices.
Assuntos
Anestesia Obstétrica/efeitos adversos , Anestesiologia/legislação & jurisprudência , Revisão da Utilização de Seguros , Responsabilidade Legal , Imperícia/legislação & jurisprudência , Adulto , Anestesia por Condução , Anestesiologia/métodos , Lesões Encefálicas/etiologia , Bases de Dados Factuais , Feminino , Humanos , Morte Materna , Gravidez , Medição de Risco , Traumatismos da Medula Espinal/etiologia , Adulto JovemRESUMO
BACKGROUND: The administration of uterotonic agents during cesarean delivery is highly variable. The authors hypothesized a "rule of threes" algorithm, featuring oxytocin 3 IU, timed uterine tone evaluations, and a systematic approach to alternative uterotonic agents, would reduce the oxytocin dose required to obtain adequate uterine tone. METHODS: Sixty women undergoing elective cesarean delivery were randomized to receive a low-dose bolus or continuous infusion of oxytocin. To blind participants, the rule group simultaneously received intravenous oxytocin (3 IU/3 ml) and a "wide-open" infusion of 0.9% normal saline (500 ml); the standard care group received intravenous 0.9% normal saline (3 ml) and a "wide-open" infusion of oxytocin (30 IU in 0.9% normal saline/500 ml). Uterine tone was assessed at 3, 6, 9, and 12 min, and if inadequate, additional uterotonic agents were administered. Uterine tone, total dose and timing of uterotonic agent use, maternal hemodynamics, side effects, and blood loss were recorded. RESULTS: Adequate uterine tone was achieved with lower oxytocin doses in the rule versus standard care group (mean, 4.0 vs. 8.4 IU; point estimate of the difference, 4.4 ± 1.0 IU; 95% CI, 2.60 to 6.15; P < 0.0001). No additional oxytocin or alternative uterotonic agents were needed in either group after 6 min. No differences in the uterine tone, maternal hemodynamics, side effects, or blood loss were observed. CONCLUSION: A "rule of threes" algorithm using oxytocin 3 IU results in lower oxytocin doses when compared with continuous-infusion oxytocin in women undergoing elective cesarean delivery.
Assuntos
Algoritmos , Cesárea/métodos , Procedimentos Cirúrgicos Eletivos/métodos , Ocitocina/administração & dosagem , Contração Uterina/efeitos dos fármacos , Adulto , Método Duplo-Cego , Feminino , Humanos , Infusões Intravenosas , Gravidez , Estudos Prospectivos , Contração Uterina/fisiologiaRESUMO
Objective: Survival analysis is widely utilized in healthcare to predict the timing of disease onset. Traditional methods of survival analysis are usually based on Cox Proportional Hazards model and assume proportional risk for all subjects. However, this assumption is rarely true for most diseases, as the underlying factors have complex, non-linear, and time-varying relationships. This concern is especially relevant for pregnancy, where the risk for pregnancy-related complications, such as preeclampsia, varies across gestation. Recently, deep learning survival models have shown promise in addressing the limitations of classical models, as the novel models allow for non-proportional risk handling, capturing nonlinear relationships, and navigating complex temporal dynamics. Methods: We present a methodology to model the temporal risk of preeclampsia during pregnancy and investigate the associated clinical risk factors. We utilized a retrospective dataset including 66,425 pregnant individuals who delivered in two tertiary care centers from 2015-2023. We modeled the preeclampsia risk by modifying DeepHit, a deep survival model, which leverages neural network architecture to capture time-varying relationships between covariates in pregnancy. We applied time series k-means clustering to DeepHit's normalized output and investigated interpretability using Shapley values. Results: We demonstrate that DeepHit can effectively handle high-dimensional data and evolving risk hazards over time with performance similar to the Cox Proportional Hazards model, achieving an area under the curve (AUC) of 0.78 for both models. The deep survival model outperformed traditional methodology by identifying time-varied risk trajectories for preeclampsia, providing insights for early and individualized intervention. K-means clustering resulted in patients delineating into low-risk, early-onset, and late-onset preeclampsia groups- notably, each of those has distinct risk factors. Conclusion: This work demonstrates a novel application of deep survival analysis in time-varying prediction of preeclampsia risk. Our results highlight the advantage of deep survival models compared to Cox Proportional Hazards models in providing personalized risk trajectory and demonstrating the potential of deep survival models to generate interpretable and meaningful clinical applications in medicine.
RESUMO
BACKGROUND: Preeclampsia, a pregnancy-specific condition associated with new-onset hypertension after 20-weeks gestation, is a leading cause of maternal and neonatal morbidity and mortality. Predictive tools to understand which individuals are most at risk are needed. METHODS: We identified a cohort of N=1125 pregnant individuals who delivered between May 2015 and May 2022 at Mass General Brigham Hospitals with available electronic health record data and linked genetic data. Using clinical electronic health record data and systolic blood pressure polygenic risk scores derived from a large genome-wide association study, we developed machine learning (XGBoost) and logistic regression models to predict preeclampsia risk. RESULTS: Pregnant individuals with a systolic blood pressure polygenic risk score in the top quartile had higher blood pressures throughout pregnancy compared with patients within the lowest quartile systolic blood pressure polygenic risk score. In the first trimester, the most predictive model was XGBoost, with an area under the curve of 0.74. In late pregnancy, with data obtained up to the delivery admission, the best-performing model was XGBoost using clinical variables, which achieved an area under the curve of 0.91. Adding the systolic blood pressure polygenic risk score to the models did not improve the performance significantly based on De Long test comparing the area under the curve of models with and without the polygenic score. CONCLUSIONS: Integrating clinical factors into predictive models can inform personalized preeclampsia risk and achieve higher predictive power than the current practice. In the future, personalized tools can be implemented to identify high-risk patients for preventative therapies and timely intervention to improve adverse maternal and neonatal outcomes.
Assuntos
Pré-Eclâmpsia , Feminino , Recém-Nascido , Gravidez , Humanos , Pré-Eclâmpsia/diagnóstico , Pré-Eclâmpsia/epidemiologia , Pré-Eclâmpsia/genética , Estratificação de Risco Genético , Estudo de Associação Genômica Ampla , Valor Preditivo dos Testes , Aprendizado de Máquina , Fatores de RiscoRESUMO
Importance: There are limited data on the association of sex with the incidence of postoperative atrial fibrillation (poAF) and subsequent long-term mortality after cardiac surgery. Objective: To evaluate whether the incidence of poAF and associated long-term mortality after cardiac surgery differ by sex. Design, Setting, and Participants: This retrospective cohort study was conducted at 2 tertiary care centers in Massachusetts from January 1, 2002, until October 1, 2016, with follow-up until December 1, 2022. Adult (aged >20 years) women and men undergoing coronary artery bypass graft surgery, aortic valve surgery, mitral valve surgery, and combined procedures with cardiopulmonary bypass were examined using medical records. Patients who had data on poAF were included in data analyses. Exposures: Sex and poAF. Main Outcomes and Measures: Primary outcomes were the incidence of poAF and all-cause mortality. poAF was defined as any atrial fibrillation detected on electrocardiogram (EKG) during the index hospitalization in patients presenting for surgery in normal sinus rhythm. Data on poAF were obtained from EKG reports and supplemented by information from the Society of Thoracic Surgeons Adult Cardiac Surgery Database. All-cause mortality was assessed via hospital records. The hypotheses were formulated prior to data analysis. Results: Among 21â¯568 patients with poAF data (mean [SD] age, 66.5 [12.4] years), 2694 of 6601 women (40.8%) and 5805 of 14â¯967 men (38.8%) developed poAF. In a multivariable logistic regression model, women had lower risk of poAF (odds ratio [OR], 0.85; 95% CI, 0.79-0.91; P < .001). During the follow-up study period, 1294 women (50.4%) and 2376 men (48.9%) in the poAF group as well as 1273 women (49.6%) and 2484 men (51.1%) in the non-poAF group died. Cox proportional hazards analysis found that the association between poAF and mortality was significantly moderated (ie, effect modified) by sex. Compared with same-sex individuals without poAF, men with poAF had a 17% higher mortality hazard (hazard ratio [HR], 1.17; 95% CI, 1.11-1.25; P < .001), and women with poAF had a 31% higher mortality hazard (HR, 1.31; 95% CI, 1.21-1.42; P < .001). Conclusions and Relevance: In this retrospective cohort study of 21â¯568 patients who underwent cardiac surgery, women were less likely to develop poAF than men when controlling for other relevant characteristics; however, women who did develop poAF had a higher risk of long-term mortality than men who developed poAF. This observed elevated risk calls for a tailored approach to perioperative care in women undergoing cardiac surgery.
Assuntos
Fibrilação Atrial , Procedimentos Cirúrgicos Cardíacos , Complicações Pós-Operatórias , Humanos , Fibrilação Atrial/epidemiologia , Fibrilação Atrial/etiologia , Fibrilação Atrial/mortalidade , Masculino , Feminino , Estudos Retrospectivos , Pessoa de Meia-Idade , Idoso , Procedimentos Cirúrgicos Cardíacos/efeitos adversos , Procedimentos Cirúrgicos Cardíacos/mortalidade , Fatores Sexuais , Complicações Pós-Operatórias/mortalidade , Complicações Pós-Operatórias/epidemiologia , Incidência , Fatores de Risco , Massachusetts/epidemiologiaRESUMO
OBJECTIVE: To evaluate the performance characteristics of existing screening tools for the prediction of sepsis during antepartum and postpartum readmissions. METHODS: This was a case-control study using electronic health record data obtained between 2016 and 2021 from 67 hospitals for antepartum sepsis admissions and 71 hospitals for postpartum readmissions up to 42 days. Patients in the sepsis case group were matched in a 1:4 ratio to a comparison cohort of patients without sepsis admitted antepartum or postpartum. The following screening criteria were evaluated: the CMQCC (California Maternal Quality Care Collaborative) initial sepsis screen, the non-pregnancy-adjusted SIRS (Systemic Inflammatory Response Syndrome), the MEWC (Maternal Early Warning Criteria), UKOSS (United Kingdom Obstetric Surveillance System) obstetric SIRS, and the MEWT (Maternal Early Warning Trigger Tool). Time periods were divided into early pregnancy (less than 20 weeks of gestation), more than 20 weeks of gestation, early postpartum (less than 3 days postpartum), and late postpartum through 42 days. False-positive screening rates, C-statistics, sensitivity, and specificity were reported for each overall screening tool and each individual criterion. RESULTS: We identified 525 patients with sepsis during an antepartum hospitalization and 728 patients with sepsis during a postpartum readmission. For early pregnancy and more than 3 days postpartum, non-pregnancy-adjusted SIRS had the highest C-statistics (0.78 and 0.83, respectively). For more than 20 weeks of gestation and less than 3 days postpartum, the pregnancy-adjusted sepsis screening tools (CMQCC and UKOSS) had the highest C-statistics (0.87-0.94). The MEWC maintained the highest sensitivity rates during all time periods (81.9-94.4%) but also had the highest false-positive rates (30.4-63.9%). The pregnancy-adjusted sepsis screening tools (CMQCC, UKOSS) had the lowest false-positive rates in all time periods (3.9-10.1%). All tools had the lowest C-statistics in the periods of less than 20 weeks of gestation and more than 3 days postpartum. CONCLUSION: For admissions early in pregnancy and more than 3 days postpartum, non-pregnancy-adjusted sepsis screening tools performed better than pregnancy-adjusted tools. From 20 weeks of gestation through up to 3 days postpartum, using a pregnancy-adjusted sepsis screening tool increased sensitivity and minimized false-positive rates. The overall false-positive rate remained high.
Assuntos
Infecção Puerperal , Sepse , Gravidez , Feminino , Humanos , Estudos de Casos e Controles , Período Pós-Parto , Hospitalização , Sepse/diagnóstico , Sepse/epidemiologia , Síndrome de Resposta Inflamatória Sistêmica , Estudos RetrospectivosRESUMO
OBJECTIVE: To evaluate the screening performance characteristics of existing tools for the diagnosis of sepsis during delivery admissions. METHODS: This was a case-control study using electronic health record data, including vital signs and laboratory results, for all delivery admissions of patients with sepsis from 59 nationally distributed hospitals. Patients with sepsis were matched by gestational age at delivery in a 1:4 ratio with patients without sepsis to create a comparison group. Patients with chorioamnionitis and sepsis were compared with a complete cohort of patients with chorioamnionitis without sepsis. Multiple screening criteria for sepsis were evaluated: the CMQCC (California Maternal Quality Care Collaborative), SIRS (Systemic Inflammatory Response Syndrome), the MEWC (the Maternal Early Warning Criteria), UKOSS (United Kingdom Obstetric Surveillance System), and the MEWT (Maternal Early Warning Trigger Tool). Sensitivity, false-positive rates, and C-statistics were reported for each screening tool. Analyses were stratified into cohort 1, which excluded patients with chorioamnionitis-endometritis, and cohort 2, which included those patients. RESULTS: Delivery admissions at 59 hospitals were extracted for patients with sepsis. Cohort 1 comprised 647 patients with sepsis, including 228 with end-organ injury, matched with a control group of 2,588 patients without sepsis. Cohort 2 comprised 14,591 patients with chorioamnionitis-endometritis, of whom 1,049 had sepsis and 238 had end-organ injury. In cohort 1, the CMQCC and the UKOSS pregnancy-adjusted criteria had the lowest false-positive rates (6.9% and 9.6%, respectively) and the highest C-statistics (0.92 and 0.91, respectively). Although other screening criteria, such as SIRS and the MEWC, had similar sensitivities, it was at the cost of much higher false-positive rates (21.3% and 38.3%, respectively). In cohort 2, including all patients with chorioamnionitis-endometritis, the highest C-statistics were again for the CMQCC (0.67) and UKOSS (0.64). All screening tools had high false-positive rates, but the false-positive rates for the CMQCC and UKOSS were substantially lower than those for SIRS and the MEWC. CONCLUSION: During delivery admissions, the CMQCC and UKOSS pregnancy-adjusted screening criteria have the lowest false-positive results while maintaining greater than 90% sensitivity rates. Performance of all screening tools was degraded in the setting of chorioamnionitis-endometritis.
Assuntos
Corioamnionite , Endometrite , Sepse , Gravidez , Feminino , Humanos , Corioamnionite/diagnóstico , Corioamnionite/epidemiologia , Estudos de Casos e Controles , Estudos Retrospectivos , Sepse/diagnóstico , Síndrome de Resposta Inflamatória SistêmicaRESUMO
Healthcare artificial intelligence (AI) holds the potential to increase patient safety, augment efficiency and improve patient outcomes, yet research is often limited by data access, cohort curation, and tools for analysis. Collection and translation of electronic health record data, live data, and real-time high-resolution device data can be challenging and time-consuming. The development of clinically relevant AI tools requires overcoming challenges in data acquisition, scarce hospital resources, and requirements for data governance. These bottlenecks may result in resource-heavy needs and long delays in research and development of AI systems. We present a system and methodology to accelerate data acquisition, dataset development and analysis, and AI model development. We created an interactive platform that relies on a scalable microservice architecture. This system can ingest 15,000 patient records per hour, where each record represents thousands of multimodal measurements, text notes, and high-resolution data. Collectively, these records can approach a terabyte of data. The platform can further perform cohort generation and preliminary dataset analysis in 2-5 minutes. As a result, multiple users can collaborate simultaneously to iterate on datasets and models in real time. We anticipate that this approach will accelerate clinical AI model development, and, in the long run, meaningfully improve healthcare delivery.
Assuntos
Inteligência Artificial , Neurofibromina 2 , Humanos , Atenção à Saúde , Pesquisa sobre Serviços de Saúde , HospitaisRESUMO
Purpose of Review: The purpose of this review is to summarize the current research and critically examine artificial intelligence (AI) technologies and their applicability to the daily practice of anesthesiologists. Recent Findings: Novel AI tools are developed using data from electronic health records, imaging, waveforms, clinical notes, and wearables. These tools can accurately predict the perioperative risk for adverse outcomes, the need for blood transfusion, and the risk of difficult intubation. Intraoperatively, AI models can assist with technical skill augmentation, patient monitoring, and management. Postoperatively, AI technology can aid in preventing complications and discharge planning. While further prospective validation is needed, these early applications demonstrate promise in every area of perioperative care. Summary: The practice of anesthesiology is at a precipice fueled by technological innovation. The clinical AI implementation would enable personalized and safer patient care by offering actionable insights from the wealth of perioperative data.
RESUMO
Background: Preeclampsia, a pregnancy-specific condition associated with new-onset hypertension after 20 weeks gestation, is a leading cause of maternal and neonatal morbidity and mortality. Predictive tools to understand which individuals are most at risk are needed. Methods: We identified a cohort of N=1,125 pregnant individuals who delivered between 05/2015-05/2022 at Mass General Brigham hospitals with available electronic health record (EHR) data and linked genetic data. Using clinical EHR data and systolic blood pressure polygenic risk scores (SBP PRS) derived from a large genome-wide association study, we developed machine learning (xgboost) and linear regression models to predict preeclampsia risk. Results: Pregnant individuals with an SBP PRS in the top quartile had higher blood pressures throughout pregnancy compared to patients within the lowest quartile SBP PRS. In the first trimester, the most predictive model was xgboost, with an area under the curve (AUC) of 0.73. Adding the SBP PRS to the models improved the performance only of the linear regression model from AUC 0.70 to 0.71; the predictive power of other models remained unchanged. In late pregnancy, with data obtained up to the delivery admission, the best performing model was xgboost using clinical variables, which achieved an AUC of 0.91. Conclusions: Integrating clinical and genetic factors into predictive models can inform personalized preeclampsia risk and achieve higher predictive power than the current practice. In the future, personalized tools can be implemented in clinical practice to identify high-risk patients for preventative therapies and timely intervention to improve adverse maternal and neonatal outcomes.
RESUMO
BACKGROUND: Current guidelines recommend prophylactic vasopressor administration during spinal anesthesia for cesarean delivery to maintain intraoperative blood pressure above 90% of the baseline value. We sought to determine the optimum baseline mean arterial pressure (MAP) reading to guide the management of spinal hypotension. METHODS: We performed a secondary analysis of data collected from normotensive patients presenting for elective cesarean delivery in a tertiary care institution from October 2018 to August 2020. We compared the magnitude of hypotension in patients who reported nausea versus those who did not, using a case-control design. Baseline MAPs at last office visit, morning of surgery, or operating room (pre-spinal) were determined. We calculated the duration and degree of hypotension using the area under the curve (AUC) when the MAP of the respective patient was below 90% of each baseline. RESULTS: The patients who experienced nausea (n=45) had longer and more profound periods of hypotension than those who did not develop nausea (n=240). A comparison of AUC using MAP baseline at the last office visit or on the morning of surgery showed a statistically significant between-group difference, P=0.02, and P=0.005, respectively, and no significant between-group difference when 90% of the MAP baseline in the operating room was used. CONCLUSIONS: Patients had the highest preoperative MAP in the operating room and the AUC was similar for those with and without nausea when the pre-spinal MAP baseline was used. Therefore, maintaining higher intraoperative blood pressure using individual pre-spinal MAP as baseline should reduce intraoperative maternal nausea.
RESUMO
Many areas of medicine would benefit from deeper, more accurate phenotyping, but there are limited approaches for phenotyping using clinical notes without substantial annotated data. Large language models (LLMs) have demonstrated immense potential to adapt to novel tasks with no additional training by specifying task-specific i nstructions. We investigated the per-formance of a publicly available LLM, Flan-T5, in phenotyping patients with postpartum hemorrhage (PPH) using discharge notes from electronic health records ( n =271,081). The language model achieved strong performance in extracting 24 granular concepts associated with PPH. Identifying these granular concepts accurately allowed the development of inter-pretable, complex phenotypes and subtypes. The Flan-T5 model achieved high fidelity in phenotyping PPH (positive predictive value of 0.95), identifying 47% more patients with this complication compared to the current standard of using claims codes. This LLM pipeline can be used reliably for subtyping PPH and outperformed a claims-based approach on the three most common PPH subtypes associated with uterine atony, abnormal placentation, and obstetric trauma. The advantage of this approach to subtyping is its interpretability, as each concept contributing to the subtype determination can be evaluated. Moreover, as definitions may change over time due to new guidelines, using granular concepts to create complex phenotypes enables prompt and efficient updating of the algorithm. Using this lan-guage modelling approach enables rapid phenotyping without the need for any manually annotated training data across multiple clinical use cases.
RESUMO
Many areas of medicine would benefit from deeper, more accurate phenotyping, but there are limited approaches for phenotyping using clinical notes without substantial annotated data. Large language models (LLMs) have demonstrated immense potential to adapt to novel tasks with no additional training by specifying task-specific instructions. Here we report the performance of a publicly available LLM, Flan-T5, in phenotyping patients with postpartum hemorrhage (PPH) using discharge notes from electronic health records (n = 271,081). The language model achieves strong performance in extracting 24 granular concepts associated with PPH. Identifying these granular concepts accurately allows the development of interpretable, complex phenotypes and subtypes. The Flan-T5 model achieves high fidelity in phenotyping PPH (positive predictive value of 0.95), identifying 47% more patients with this complication compared to the current standard of using claims codes. This LLM pipeline can be used reliably for subtyping PPH and outperforms a claims-based approach on the three most common PPH subtypes associated with uterine atony, abnormal placentation, and obstetric trauma. The advantage of this approach to subtyping is its interpretability, as each concept contributing to the subtype determination can be evaluated. Moreover, as definitions may change over time due to new guidelines, using granular concepts to create complex phenotypes enables prompt and efficient updating of the algorithm. Using this language modelling approach enables rapid phenotyping without the need for any manually annotated training data across multiple clinical use cases.
RESUMO
Background: Preeclampsia is a pregnancy-specific disease characterized by new onset hypertension after 20 weeks of gestation that affects 2-8% of all pregnancies and contributes to up to 26% of maternal deaths. Despite extensive clinical research, current predictive tools fail to identify up to 66% of patients who will develop preeclampsia. We sought to develop a tool to longitudinally predict preeclampsia risk. Methods: In this retrospective model development and validation study, we examined a large cohort of patients who delivered at six community and two tertiary care hospitals in the New England region between 02/2015 and 06/2023. We used sociodemographic, clinical diagnoses, family history, laboratory, and vital signs data. We developed eight datasets at 14, 20, 24, 28, 32, 36, 39 weeks gestation and at the hospital admission for delivery. We created linear regression, random forest, xgboost, and deep neural networks to develop multiple models and compared their performance. We used Shapley values to investigate the global and local explainability of the models and the relationships between the predictive variables. Findings: Our study population (N=120,752) had an incidence of preeclampsia of 5.7% (N=6,920). The performance of the models as measured using the area under the curve, AUC, was in the range 0.73-0.91, which was externally validated. The relationships between some of the variables were complex and non-linear; in addition, the relative significance of the predictors varied over the pregnancy. Compared to the current standard of care for preeclampsia risk stratification in the first trimester, our model would allow 48.6% more at-risk patients to be identified. Interpretation: Our novel preeclampsia prediction tool would allow clinicians to identify patients at risk early and provide personalized predictions, as well as longitudinal predictions throughout pregnancy. Funding: National Institutes of Health, Anesthesia Patient Safety Foundation. RESEARCH IN CONTEXT: Evidence before this study: Current tools for the prediction of preeclampsia are lacking as they fail to identify up to 66% of the patients who develop preeclampsia. We searched PubMed, MEDLINE, and the Web of Science from database inception to May 1, 2023, using the keywords "deep learning", "machine learning", "preeclampsia", "artificial intelligence", "pregnancy complications", and "predictive models". We identified 13 studies that employed machine learning to develop prediction models for preeclampsia risk based on clinical variables. Among these studies, six included biomarkers such as serum placental growth factor, pregnancy-associated plasma protein A, and uterine artery pulsatility index, which are not routinely available in our clinical practice; two studies were in diverse cohorts of more than 100 000 patients, and two studies developed longitudinal predictions using medical records data. However, most studies have limited depth, concerns about data leakage, overfitting, or lack of generalizability.Added value of this study: We developed a comprehensive longitudinal predictive tool based on routine clinical data that can be used throughout pregnancy to predict the risk of preeclampsia. We tested multiple types of predictive models, including machine learning and deep learning models, and demonstrated high predictive power. We investigated the changes over different time points of individual and group variables and found previously known and novel relationships between variables such as red blood cell count and preeclampsia risk.Implications of all the available evidence: Longitudinal prediction of preeclampsia using machine learning can be achieved with high performance. Implementation of an accurate predictive tool within the electronic health records can aid clinical care and identify patients at heightened risk who would benefit from aspirin prophylaxis, increased surveillance, early diagnosis, and escalation in care. These results highlight the potential of using artificial intelligence in clinical decision support, with the ultimate goal of reducing iatrogenic preterm birth and improving perinatal care.
RESUMO
OBJECTIVE: To evaluate and understand pregnant patients' perspectives on the implementation of artificial intelligence (AI) in clinical care with a focus on opportunities to improve healthcare technologies and healthcare delivery. MATERIALS AND METHODS: We developed an anonymous survey and enrolled patients presenting to the labor and delivery unit at a tertiary care center September 2019-June 2020. We investigated the role and interplay of patient demographic factors, healthcare literacy, understanding of AI, comfort levels with various AI scenarios, and preferences for AI use in clinical care. RESULTS: Of the 349 parturients, 57.6% were between the ages of 25-34 years, 90.1% reported college or graduate education and 69.2% believed the benefits of AI use in clinical care outweighed the risks. Cluster analysis revealed 2 distinct groups: patients more comfortable with clinical AI use (Pro-AI) and those who preferred physician presence (AI-Cautious). Pro-AI patients had a higher degree of education, were more knowledgeable about AI use in their daily lives and saw AI use as a significant advancement in medicine. AI-Cautious patients reported a lack of human qualities and low trust in the technology as detriments to AI use. DISCUSSION: Patient trust and the preservation of the human physician-patient relationship are critical in moving forward with AI implementation in healthcare. Pregnant individuals are cautiously optimistic about AI use in their care. CONCLUSION: Our findings provide insights into the status of AI use in perinatal care and provide a platform for driving patient-centered innovations.