WAVES - The Lucile Packard Children's Hospital Pediatric Physiological Waveforms Dataset

WAVES is a large, single-center dataset comprising 9 years of high-frequency physiological waveform data from patients in intensive and acute care units at a large academic, pediatric medical center. The data comprise approximately 10.6 million hours of 1 to 20 concurrent waveforms over approximately 50,364 distinct patient encounters. The data have been de-identified, cleaned, and organized to facilitate research. Initial analyses demonstrate the potential of the data for clinical applications such as non-invasive blood pressure monitoring and methodological applications such as waveform-agnostic data imputation. WAVES is the largest pediatric-focused and second largest physiological waveform dataset available for research.

Association of Common Pediatric Surgeries With New Onset Chronic Pain in Patients 0-21 Years of Age in the United States

Chronic pain (CP) is a major public health issue. While new onset CP is known to occur frequently after some pediatric surgeries, its incidence after the most common pediatric surgeries is unknown. This retrospective cohort study used insurance claims data from 2002 to 2017 for patients 0 to 21 years of age. The primary outcome was CP 90 to 365 days after each of the 20 most frequent surgeries in 5 age categories (identified using CP ICD codes). Multivariable logistic regression identified surgeries and risk factors associated with CP after surgery. A total of 424,590 surgical patients aged 0 to 21 were included, 22,361 of whom developed CP in the 90 to 365 days after surgery. The incidences of CP after surgery were: 1.1% in age group 0 to 1 years; 3.0% in 2 to 5 years; 5.6% in 6 to 11 years; 10.1% in 12 to 18 years; 9.9% in 19 to 21 years. Some surgeries and patient variables were associated with CP. Approximately 1 in 10 adolescents who underwent the most common surgeries developed CP, as did a striking percentage of children in other age groups. Given the long-term consequences of CP, resources should be allocated toward identification of high-risk pediatric patients and strategies to prevent CP after surgery. PERSPECTIVE: This study identifies the incidences of and risk factors for chronic pain after common surgeries in patients 0 to 21 years of age. Our findings suggest that resources should be allocated toward the identification of high-risk pediatric patients and strategies to prevent CP after surgery.

Wireless Body Area Network Control Policies for Energy-Efficient Health Monitoring

Wireless body area networks (WBANs) have strong potential in the field of health monitoring. However, the energy consumption required for accurate monitoring determines the time between battery charges of the wearable sensors, which is a key performance factor (and can be critical in the case of implantable devices). In this paper, we study the inherent trade-off between the power consumption of the sensors and the probability of misclassifying a patient's health state. We formulate this trade-off as a dynamic problem, in which at each step, we can choose to activate a subset of sensors that provide noisy measurements of the patient's health state. We assume that the (unknown) health state follows a Markov chain, so our problem is formulated as a partially observable Markov decision problem (POMDP). We show that all the past measurements can be summarized as a belief state on the true health state of the patient, which allows tackling the POMDP problem as an MDP on the belief state. Then, we empirically study the performance of a greedy one-step look-ahead policy compared to the optimal policy obtained by solving the dynamic program. For that purpose, we use an open-source Continuous Glucose Monitoring (CGM) dataset of 232 patients over six months and extract the transition matrix and sensor accuracies from the data. We find that the greedy policy saves ≈50% of the energy costs while reducing the misclassification costs by less than 2% compared to the most accurate policy possible that always activates all sensors. Our sensitivity analysis reveals that the greedy policy remains nearly optimal across different cost parameters and a varying number of sensors. The results also have practical importance, because while the optimal policy is too complicated, a greedy one-step look-ahead policy can be easily implemented in WBAN systems.

Prediction of Prolonged Opioid Use After Surgery in Adolescents: Insights From Machine Learning

BACKGROUND

Long-term opioid use has negative health care consequences. Patients who undergo surgery are at risk for prolonged opioid use after surgery (POUS). While risk factors have been previously identified, no methods currently exist to determine higher-risk patients. We assessed the ability of a variety of machine-learning algorithms to predict adolescents at risk of POUS and to identify factors associated with this risk.

METHODS

A retrospective cohort study was conducted using a national insurance claims database of adolescents aged 12-21 years who underwent 1 of 1297 surgeries, with general anesthesia, from January 1, 2011 to December 30, 2017. Logistic regression with an L2 penalty and with a logistic regression with an L1 lasso (Lasso) penalty, random forests, gradient boosting machines, and extreme gradient boosted models were trained using patient and provider characteristics to predict POUS (≥1 opioid prescription fill within 90-180 days after surgery) risk. Predictive capabilities were assessed using the area under the receiver-operating characteristic curve (AUC)/C-statistic, mean average precision (MAP); individual decision thresholds were compared using sensitivity, specificity, Youden Index, F1 score, and number needed to evaluate. The variables most strongly associated with POUS risk were identified using permutation importance.

RESULTS

Of 186,493 eligible patient surgical visits, 8410 (4.51%) had POUS. The top-performing algorithm achieved an overall AUC of 0.711 (95% confidence interval [CI], 0.699-0.723) and significantly higher AUCs for certain surgeries (eg, 0.823 for spinal fusion surgery and 0.812 for dental surgery). The variables with the strongest association with POUS were the days' supply of opioids and oral morphine milligram equivalents of opioids in the year before surgery.

CONCLUSIONS

Machine-learning models to predict POUS risk among adolescents show modest to strong results for different surgeries and reveal variables associated with higher risk. These results may inform health care system-specific identification of patients at higher risk for POUS and drive development of preventative measures.

Incidence of and Factors Associated With Prolonged and Persistent Postoperative Opioid Use in Children 0-18 Years of Age

BACKGROUND

Long-term opioid use has negative health care consequences. Opioid-naïve adults are at risk for prolonged and persistent opioid use after surgery. While these outcomes have been examined in some adolescent and teenage populations, little is known about the risk of prolonged and persistent postoperative opioid use after common surgeries compared to children who do not undergo surgery and factors associated with these issues among pediatric surgical patients of all ages.

METHODS

Using a national administrative claims database, we identified 175,878 surgical visits by opioid-naïve children aged ≤18 years who underwent ≥1 of the 20 most common surgeries from each of 4 age groups between December 31, 2002, and December 30, 2017, and who filled a perioperative opioid prescription 30 days before to 14 days after surgery. Prolonged opioid use after surgery (filling ≥1 opioid prescription 90-180 days after surgery) was compared to a reference sample of 1,354,909 nonsurgical patients randomly assigned a false "surgery" date. Multivariable logistic regression models were used to estimate the association of surgical procedures and 22 other variables of interest with prolonged opioid use and persistent postoperative opioid use (filling ≥60 days' supply of opioids 90-365 days after surgery) for each age group.

RESULTS

Prolonged opioid use after surgery occurred in 0.77%, 0.76%, 1.00%, and 3.80% of surgical patients ages 0-<2, 2-<6, 6-<12, and 12-18, respectively. It was significantly more common in surgical patients than in nonsurgical patients (ages 0-<2: odds ratio [OR] = 4.6 [95% confidence interval (CI), 3.7-5.6]; ages 2-<6: OR = 2.5 [95% CI, 2.1-2.8]; ages 6-<12: OR = 2.1 [95% CI, 1.9-2.4]; and ages 12-18: OR = 1.8 [95% CI, 1.7-1.9]). In the multivariable models for ages 0-<12 years, few surgical procedures and none of the other variables of interest were associated with prolonged opioid use. In the models for ages 12-18 years, 10 surgical procedures and 5 other variables of interest were associated with prolonged opioid use. Persistent postoperative opioid use occurred in <0.1% of patients in all age groups.

CONCLUSIONS

Some patient characteristics and surgeries are positively and negatively associated with prolonged opioid use in opioid-naïve children of all ages, but persistent opioid use is rare. Specific pediatric subpopulations (eg, older patients with a history of mood/personality disorder or chronic pain) may be at markedly higher risk.

On stability of queueing networks with job deadlines

In this paper, we consider a single-server queue with stationary input, where each job joining the queue has an associated deadline. The deadline is a time constraint on job sojourn time and may be finite or infinite. If the job does not complete service before its deadline expires, it abandons the queue and the partial service it may have received up to that point is wasted. When the queue operates under a first-come-first served discipline, we establish conditions under which the actual workload process—that is, the work the server eventually processes—is unstable, weakly stable, and strongly stable. An interesting phenomenon observed is that in a nontrivial portion of the parameter space, the queue is weakly stable, but not strongly stable. We also indicate how our results apply to other nonidling service disciplines. We finally extend the results for a single node to acyclic (feed-forward) networks of queues with either per-queue or network-wide deadlines.

On queues with periodic inputs

We consider a single-server queue with a periodic and ergodic input. It is shown that if the traffic intensity is less than 1, then the waiting time process is asymptotically periodic. Limit theorems associated with the asymptotic behavior of the queue are also proven. The results are then extended to acyclic networks of queues with periodic inputs. Particular cases of these results had been previously obtained for a single queue with periodic Poisson arrival input process and with independent and identically distributed service times.

On closed ring queueing networks

In this paper we first study ring structured closed queueing networks with distinguishable jobs. Under assumptions of periodicity and ergodicity of the service times, essentially the most general, it is shown that the limits defining the average flows of the jobs exist almost surely, and methods for their estimation by simulation are given. However, it turns out that the values of the flows depend on the initial positions of the jobs, due to the emergence of distinct persistent blocking modes. The effect of these modes on the behavior of general networks with queueing loops is examined.

For independent and identically distributed service times, conditions are specified for the network to asymptotically approach a steady state at large times.

Finally, we study the special case of ring networks with indistinguishable items and stationary and ergodic service times. It is shown that as the number of jobs in the network increases towards infinity, the average circulation time converges to the maximum of the expectations of the service times.

On stability of state-dependent queues and acyclic queueing networks

We consider a single server first-come-first-served queue with a stationary and ergodic input. The service rate is a general function of the workload in the queue. We provide the necessary and sufficient conditions for the stability of the system and the asymptotic convergence of the workload process to a finite stationary process at large times. Then, we consider acyclic networks of queues in which the service rate of any queue is a function of the workloads of this and of all the preceding queues. The stability problem is again studied. The results are then extended to analogous systems with periodic inputs.

Scheduling and stability aspects of a general class of parallel processing systems

In this paper we study the following general class of concurrent processing systems. There are several different classes of processors (servers) and many identical processors within each class. There is also a continuous random flow of jobs, arriving for processing at the system. Each job needs to engage concurrently several processors from various classes in order to be processed. After acquiring the needed processors the job begins to be executed. Processing is done non-preemptively, lasts for a random amount of time, and then all the processors are released simultaneously. Each job is specified by its arrival time, its processing time, and the list of processors that it needs to access simultaneously. The random flow (sequence) of jobs has a stationary and ergodic structure. There are several possible policies for scheduling the jobs on the processors for execution; it is up to the system designer to choose the scheduling policy to achieve certain objectives.

We focus on the effect that the choice of scheduling policy has on the asymptotic behavior of the system at large times and especially on its stability, under general stationary and ergocic input flows.

An invariant distribution for the G/G/1 queueing operator

We consider the G/G/1 queue as an operator that maps inter-arrival times to inter-departure times of points, given the service times. For arbitrarily fixed statistics of service times, we are interested in the existence of distributions of inter-arrival times that induce identical distributions on the inter-departure times. In this note we prove, by construction, the existence of one of such distribution.

Queueing and scheduling in random environments

We consider a processing system, composed of several parallel queues and a processor, which operates in a time-varying environment that fluctuates between various states or modes. The service rate at each queue depends on the processor bandwidth allocated to it, as well as the environment mode. Each queue is driven by a job traffic flow, which may also depend on the environment mode. Dynamic processor scheduling policies are investigated for maximizing the system throughput, by adapting to queue backlogs and the environment mode. We show that allocating the processor bandwidth to the queues, so as to maximize the projection of the service rate vector onto a linear function of the workload vector, can keep the system stable under the maximum possible traffic load. The analysis of the system dynamics is first done under very general assumptions, addressing rate stability and flow conservation on individual traffic and environment evolution traces. The connection with stochastic stability is later discussed for stationary and ergodic traffic and environment processes. Various extensions to feed-forward networks of such nodes, the multi-processor case, etc., are also discussed. The approach advances the methodology of trace-based modelling of queueing structures. Applications of the model include bandwidth allocation in wireless channels with fluctuating interference and allocation of switching bandwidth to traffic flows in communication networks with fluctuating congestion levels.

The synchronization of Poisson processes and queueing networks with service and synchronization nodes

This paper investigates the dynamics of a synchronization node in isolation, and of networks of service and synchronization nodes. A synchronization node consists of M infinite capacity buffers, where tokens arriving on M distinct random input flows are stored (there is one buffer for each flow). Tokens are held in the buffers until one is available from each flow. When this occurs, a token is drawn from each buffer to form a group-token, which is instantaneously released as a synchronized departure. Under independent Poisson inputs, the output of a synchronization node is shown to converge weakly (and in certain cases strongly) to a Poisson process with rate equal to the minimum rate of the input flows. Hence synchronization preserves the Poisson property, as do superposition, Bernoulli sampling and M/M/1 queueing operations. We then consider networks of synchronization and exponential server nodes with Bernoulli routeing and exogenous Poisson arrivals, extending the standard Jackson network model to include synchronization nodes. It is shown that if the synchronization skeleton of the network is acyclic (i.e. no token visits any synchronization node twice although it may visit a service node repeatedly), then the distribution of the joint queue-length process of only the service nodes is product form (under standard stability conditions) and easily computable. Moreover, the network output flows converge weakly to Poisson processes. Finally, certain results for networks with finite capacity buffers are presented, and the limiting behavior of such networks as the buffer capacities become large is studied.

Personalized Monitors for Real-Time Detection of Physiological States

The authors introduce an algorithmic framework to process real-time physiological data using nonparametric Bayesian models under the context of developing and testing personalized wellness monitors. A wearable device aggregates signals from various sensors while periodically transmitting the collected data to a backend server, which builds custom user profiles based on inferred hidden Markov states. They discuss how these user profiles can be used in various contexts as proxies for fluctuating physiological states and leveraged for various longitudinal classification tasks. Using data collected in a two-week study hosted at Jaslok Hospital, the authors show how physiological changes induced by different environments with various levels of stress can be quantified by the authors' platform. To minimize the dependence on continuous connectivity with the backend server, they introduce a heuristic to enable real-time state identification using the modest processing capabilities of the wearable device.

A fuzzy reinforcement learning approach to power control in wireless transmitters

We address the issue of power-controlled shared channel access in wireless networks supporting packetized data traffic. We formulate this problem using the dynamic programming framework and present a new distributed fuzzy reinforcement learning algorithm (ACFRL-2) capable of adequately solving a class of problems to which the power control problem belongs. Our experimental results show that the algorithm converges almost deterministically to a neighborhood of optimal parameter values, as opposed to a very noisy stochastic convergence of earlier algorithms. The main tradeoff facing a transmitter is to balance its current power level with future backlog in the presence of stochastically changing interference. Simulation experiments demonstrate that the ACFRL-2 algorithm achieves significant performance gains over the standard power control approach used in CDMA2000. Such a large improvement is explained by the fact that ACFRL-2 allows transmitters to learn implicit coordination policies, which back off under stressful channel conditions as opposed to engaging in escalating "power wars."