STAN: spatio-temporal attention network for pandemic prediction using real-world evidence.

OBJECTIVE: We aim to develop a hybrid model for earlier and more accurate predictions for the number of infected cases in pandemics by (1) using patients' claims data from different counties and states that capture local disease status and medical resource utilization; (2) utilizing demographic similarity and geographical proximity between locations; and (3) integrating pandemic transmission dynamics into a deep learning model. MATERIALS AND METHODS: We proposed a spatio-temporal attention network (STAN) for pandemic prediction. It uses a graph attention network to capture spatio-temporal trends of disease dynamics and to predict the number of cases for a fixed number of days into the future. We also designed a dynamics-based loss term for enhancing long-term predictions. STAN was tested using both real-world patient claims data and COVID-19 statistics over time across US counties. RESULTS: STAN outperforms traditional epidemiological models such as susceptible-infectious-recovered (SIR), susceptible-exposed-infectious-recovered (SEIR), and deep learning models on both long-term and short-term predictions, achieving up to 87% reduction in mean squared error compared to the best baseline prediction model. CONCLUSIONS: By combining information from real-world claims data and disease case counts data, STAN can better predict disease status and medical resource utilization.

STAN: Spatio-Temporal Attention Network for Pandemic Prediction Using Real World Evidence.

OBJECTIVE: The COVID-19 pandemic has created many challenges that need immediate attention. Various epidemiological and deep learning models have been developed to predict the COVID-19 outbreak, but all have limitations that affect the accuracy and robustness of the predictions. Our method aims at addressing these limitations and making earlier and more accurate pandemic outbreak predictions by (1) using patients' EHR data from different counties and states that encode local disease status and medical resource utilization condition; (2) considering demographic similarity and geographical proximity between locations; and (3) integrating pandemic transmission dynamics into deep learning models. MATERIALS AND METHODS: We proposed a spatio-temporal attention network (STAN) for pandemic prediction. It uses an attention-based graph convolutional network to capture geographical and temporal trends and predict the number of cases for a fixed number of days into the future. We also designed a physical law-based loss term for enhancing long-term prediction. STAN was tested using both massive real-world patient data and open source COVID-19 statistics provided by Johns Hopkins university across all U.S. counties. RESULTS: STAN outperforms epidemiological modeling methods such as SIR and SEIR and deep learning models on both long-term and short-term predictions, achieving up to 87% lower mean squared error compared to the best baseline prediction model. CONCLUSIONS: By using information from real-world patient data and geographical data, STAN can better capture the disease status and medical resource utilization information and thus provides more accurate pandemic modeling. With pandemic transmission law based regularization, STAN also achieves good long-term prediction performance.

Antihypertensive treatment and blood pressure control relative to hypertension treatment guidelines.

PURPOSE: To describe initial antihypertensive management relative to important aspects of JNC7 hypertension guidelines, to identify predictors of receiving JNC7 discordant therapy, and to determine the association between receiving JNC7-concordant antihypertensive treatment and achieving blood pressure (BP) control. This study focused on aspects of the JNC7 guidelines which are consistent with other guidelines that have been published since JNC7. METHODS: EMR data from eleven multi-specialty medical groups in the US were retrospectively collected between 2008-2011. The study cohort included incident hypertensive patients who received an antihypertensive prescription during the 6-month follow-up period. Patients with existing hypertension were excluded. JNC7-concordance of the prescribed antihypertensive regimen was evaluated. Using multivariable logistic regression, we determined the association between JNC7-concordance and achieving BP control. Additionally, we determined predictors of receiving JNC7-discordant treatment. RESULTS: 14,910 incident hypertensive patients who were treated with an antihypertensive during the 6-month follow-up period were included. Overall, 79.4% patients were prescribed antihypertensive therapy concordant with JNC7; however among patients with stage 2 hypertension, the concordance was found to be 50%. BP control was achieved by 64.1% and 48.5% of patients who received JNC7-concordant and JNC7-discordant therapy, respectively. The overall adjusted odds ratio (95% CI) for BP control and JNC7-concordance was 1.53 (1.40, 1.68). The association was attenuated for cohort members with diabetes, chronic kidney disease (CKD), and stage 2 hypertension. Predictors of receiving JNC7-discordant therapy were congestive heart failure, CKD, and diabetes. CONCLUSION: JNC7-concordance is high overall, but drops substantially when JNC7 recommendations are more demanding (e.g., among patients with stage 2 hypertension and/or CKD, CHF, diabetes). Overall, patients who are prescribed an antihypertensive regimen that is JNC7-concordant are more likely to achieve BP control.

Phase 2 study of TAK-442, an oral factor Xa inhibitor, in patients following acute coronary syndrome.

TAK-442 is an oral direct factor Xa inhibitor. We sought to determine the dose-dependent effect of TAK-442 on major bleeding when added to standard treatment in stabilised patients with acute coronary syndrome (ACS). In this phase II double-blind study, 2,753 ACS patients were randomised to TAK-442 or placebo in addition to usual care using a three-stage adaptive design. Patients were randomised to placebo in all stages, but doses of TAK-442 escalated from 10 mg BID, 20 mg twice-daily (BID), or 40 mg once-daily (QD) in stage 1; to 40 mg BID, 80 mg QD, or 80 mg BID in stage 2; and to 160 mg QD or 120 mg BID in stage 3. Study drug was started 36 hours after emergent treatment of ACS and within seven days of admission, and continued for 24 weeks. The primary endpoint was incidence of TIMI (thrombolysis in myocardial infarction) major bleeding. TIMI major bleeding incidence was low, but higher with the pooled TAK-442 doses than with placebo (17 [0.9%] vs 4 [0.5%]; p=0.47), although the difference was neither significant nor dose-dependent. However, a dose response was evident when using the modified ISTH scale. The incidence of cardiovascular events was similar among TAK-442 dose groups and placebo. When administered over a wide range of doses after an ACS event, TAK-442 treatment did not result in a dose-dependent increase in TIMI major bleeding, but increased bleeding was observed when a more sensitive bleeding scale was used. There was no evidence for efficacy.

A dose-finding study with TAK-442, an oral factor Xa inhibitor, in patients undergoing elective total knee replacement surgery.

This multicentre dose-finding study compared TAK-442, an oral factor Xa inhibitor, with enoxaparin for thromboprophylaxis after knee arthroplasty. In this parallel group study, patients were randomised to oral TAK-442 (40 or 80 mg once-daily [QD] or 10, 20, 40, or 80 mg twice-daily [BID] started 6-8 hours postoperatively), which was blinded as to dose, or to open-label subcutaneous enoxaparin (30 mg BID starting 12-24 hours postoperatively) for 10 days. Treatments were continued until bilateral venography was performed (maximum of 14 days). The primary efficacy endpoint was the composite of any deep-vein thrombosis, non-fatal pulmonary embolism or all-cause mortality, while the primary safety endpoint was major bleeding. Of 1,038 patients randomised who received at least one dose of study drug, 949 completed the study and 730 (76.9%) were evaluable for the primary efficacy analysis. Recruitment into the 10 and 20 mg BID dose groups was stopped early because the incidences of the primary efficacy endpoint were significantly higher than that with enoxaparin. The primary efficacy endpoint occurred in 22.0% of patients given enoxaparin and in 39.0%, 38.4%, 23.5%, 21.4%, 26.8%, and 14.3% of those receiving TAK-442 10 mg BID, 20 mg BID, 40 mg QD, 40 mg BID, 80 mg QD, and 80 mg BID, respectively. The incidences of major and clinically relevant non-major bleeding with TAK-442 were not dose-dependent or different from that with enoxaparin. All TAK-442 doses except 10 and 20 mg BID displayed similar efficacy and safety profiles to enoxaparin.

Monitoring adverse events following yellow fever vaccination using an integrated telephone and Internet-based system.

TeleWatch is an automated telephone/Internet-based system that collects medical product adverse event reports in real-time through an algorithm driven by the patient. 1341 patients, who received yellow fever vaccine and were recruited through 15 travel clinics, contacted the system within 48h of vaccination and 765 (57%) made follow-up contacts. Participation rates were higher among females and persons older than 60 years of age. TeleWatch can be expanded for use in large campaigns involving influenza or other vaccines.

Rapid titration of carvedilol in patients with congestive heart failure: a randomized trial of automated telemedicine versus frequent outpatient clinic visits.

BACKGROUND: Because of potential side effects and logistical difficulty of titrating medications, outpatients with congestive heart failure rarely receive appropriate doses of carvedilol or other beta-blockers. To address these obstacles, we studied if an automated telemedicine system named TeleWatch (TW) could facilitate carvedilol titration in outpatients with left ventricular systolic dysfunction. METHODS: Forty-nine patients with New York Heart Association class II and III left ventricular systolic dysfunction, who were tolerating appropriate afterload-reducing therapy and not receiving beta-blockers, were enrolled into a 3-month study. Patients were randomized to have clinic-only (CO) (n = 24) carvedilol titration or titrations which combined clinic visits with TW monitoring (n = 25). All patients were seen in clinic biweekly, and those in the TW group were remotely monitored daily. Using a predefined algorithm, patients in the CO and TW groups were eligible for carvedilol titration on a biweekly or weekly basis, respectively, by physicians blinded to group assignment. RESULTS: There was no statistical difference in the mean final daily dose of carvedilol between the CO and TW groups (39.4 vs 36.2 mg/d, P = .52). Because remote telemedicine titrations were as successful as titrations in the clinic, the time to reach the final dose of carvedilol was significantly shorter in the TW group (33.6 vs 63.7 days, P < .0001). There were 5 serious adverse events in the study, 4 of which were in the TW group (P = .29); however, TW prospectively detected 2 adverse events. CONCLUSIONS: Remote monitoring with an automated telemedicine system can successfully facilitate titration of carvedilol in outpatients with New York Heart Association class II and III congestive heart failure.

Pilot study optimizing MEDLINE queries in an automated disease management telemedicine system.

Clinicians encounter many medical questions while providing outpatient medical care. A significant number of these questions can be answered using MEDLINE; however it has proven to be difficult to incorporate MEDLINE into routine clinical workflow and for clinicians to generate well constructed MEDLINE queries. This study however hypothesized that that well-constructed MEDLINE queries could be semi-automatically generated by an application named LitButton which was incorporated into the TeleWatch telemedicine system. The LitButton application was then prospectively evaluated in a pilot study by four nurse case managers (NCM) who monitored sixty-eight outpatients for three weeks. During this period the NCMs used the LitButton application sixteen times, and they subjectively reported in real-time that they obtained an answer in 56% of the cases, but that none of the successful information retrieval events resulted in a change in a patient's clinical management. The small number of LitButton events and lack of clinical impact was likely due to the fact that the LitButton function was designed to search MEDLINE for treatment related information; however the NCMs had limited medical decision making responsibilities. As a result there was a mismatch between the user's information needs and the system capabilities.