Bedside monitors are intended as a safety net in patient care, but their management in the inpatient setting is a significant patient safety concern. The low precision of vital sign alarm systems leads to clinical staff becoming desensitized to the sound of the alarm, a phenomenon known as alarm fatigue. Alarm fatigue has been shown to increase response time to alarms or result in alarms being ignored altogether and has negative consequences for patient safety. We present methods to establish personalized thresholds for heart rate and respiratory rate alarms. These thresholds are first chosen based on patient characteristics available at the time of admission and are then adapted to incorporate vital signs observed in the first 2 hours of monitoring. We demonstrate that the adapted thresholds are similar to those chosen by clinicians for individual patients and would result in fewer alarms than the currently used age-based thresholds. Personalized vital sign alarm thresholds can help to alleviate the problem of alarm fatigue in an inpatient setting while ensuring that all critical vital signs are detected. 1846 bii Biomedical Informatics Insights Biomed Inform Insights SAGE Publications PMC6330722 6330722 6330722 30675101 10.1177/1178222618818478 Incorporating Observed Physiological Data to Personalize Pediatric Vital Sign Alarm Thresholds Poole Sarah 1 Shah Nigam 1 ✉ 1 Stanford Center for Biomedical Informatics Research, Stanford University, Stanford, CA, USA ✉ Nigam Shah, Stanford Center for Biomedical Informatics Research, Stanford University, MSOB, 1265 Welch Road, Stanford, CA 94305, USA. This study focuses on addressing alarm fatigue in a pediatric population, as the change in vital signs with age results in higher inter-patient variation compared to the adult population. 17 , 20 Research has shown that the default heart rate alarm thresholds in a pediatric hospital often fall near the 50th percentile of heart rate values observed in the population, 18 , 21 causing alarms to be triggered for over half of the observed heart rates for some patients. Our previous work used nurse-charted vital sign data to choose population-specific vital sign alarm thresholds. (B) The heart rate values that would trigger an alarm if the observed 1st and 99th percentiles of heart rate are used as alarm thresholds. The shaded regions are those that would trigger alarms. Many fewer alarms would sound under this model, and alarms would be triggered for different thresholds for each patient. This study aims to address vital sign alarm fatigue by finding personalized alarm thresholds for heart rate and respiratory rate alarms. Models are trained to find vital sign alarm thresholds on a patient-by-patient basis, rather than using patient groups. The personalized thresholds are initially built using data available at admission and are adjusted after observation of the patient’s own vital signs over a 2-hour period. A method to evaluate the resulting patient-specific alarm thresholds is also introduced. We identify alarms that had a setting chosen by a physician specifically for the patient of interest and use these alarm settings as the best available standard against which to evaluate our proposed thresholds. Methods Data Two main sources of data were used for this study. (2) V S 1 % = e − 2 . 3263 σ + μ (3) V S 99 % = e 2 . 3264 σ + μ Model structure Modeling the change in expected vital signs with age is an important component of building personalized alarm thresholds. We use loess models to capture the non-linear variation in the mean and the variance of the vital signs with age. The output of the loess models is then used as input to Bayesian additive regression tree (BART) models, along with additional demographic (age, weight, gender, ethnicity, and race) and diagnostic features (DRG and hospital department). Once the patient has had their vital signs monitored for 2 hours, we have additional data to inform the expected distribution of vital signs. This 2-hour period length was chosen to give enough time for the patient’s condition to begin to stabilize after being admitted, while still The parameters defining the posterior distribution can then be calculated using the following equations. A tilde is used to indicate the updated parameters of the distribution. (8) μ d a t a = 1 n d a t a ∑ i = 1 n d a t a V S i (9) σ d a t a 2 = 1 n n e w ∑ i = 1 n n e w ( V S i − μ d a t a ) 2 (10) α ~ = α ^ n n e w 2 (11) β ~ = β ^ + σ ^ d a t a 2 n d a t a 2 + n d a t a n 0 ( μ d a t a − μ ^ ) 2 2 ( n d a t a + n 0 ) (12) σ ~ 2 = β ~ ( α ~ − 1 ) (13) μ ~ = n d a t a μ d a t a n d a t a + n 0 + n 0 μ ^ n d a t a + n 0 We can then use equations (2) and ( 3 ) to obtain the new suggested alarm thresholds from these updated estimates of the distribution mean and variance. Note that only alarms that were actually historically triggered are part of this analysis. The dashed lines show the total number of alarms that were triggered. The alarms making up this analysis span a 3.5-year period. Discussion The aim of this study was to find personalized vital sign alarm thresholds for hospitalized pediatric patients without requiring input from clinical staff. One of the challenges to achieving this aim was determining an appropriate target for these thresholds. Due to the lack of a large set of labeled alarms to use in training a model, we needed to use the patient’s observed vital signs to choose targets for the threshold values. Conclusions In conclusion, this study describes methods to choose personalized heart rate and respiratory rate alarm thresholds for pediatric inpatients. These thresholds are initially chosen based on patient characteristics available at admission and are adapted to incorporate vital signs observed over the first 2 hours of vital sign monitoring. The resulting adapted thresholds are similar to physician-selected thresholds and result in fewer alarms than the currently used thresholds.

Epilepsy affects over 50 million people worldwide, yet automated seizure detection systems either achieve moderate sensitivity with excessive false alarms or rely on uninterpretable deep networks. This study presents a patient-independent EEG-based seizure detection framework that achieved zero false alarms in 24 h with 95% sensitivity in a retrospective evaluation on a CHB-MIT pediatric cohort (<i>n</i> = 6 seizure-positive patients). The pipeline extracts 27 time-, frequency-, and nonlinear-domain features from 5 s windows and trains five ensemble classifiers (XGBoost, CatBoost, LightGBM, Extra Trees, Random Forest) using strict leave-one-subject-out cross-validation. All models achieved segment-level AUC ≥ 0.99. Under zero-false-alarm constraints, XGBoost attained perfect specificity with 0.922 sensitivity. SHAP and LIME analyses suggested candidate EEG biomarkers that appear consistent with known ictal signatures, including temporo-parietal theta-band power, amplitude variability (IQR, RMS), and Hjorth activity. External validation on the Siena Scalp EEG Database (12 adult patients, 37 seizures) demonstrated cross-dataset generalization with 95% event-level sensitivity (Extra Trees) and AUC of 0.86 (Random Forest). Temporal lobe channels dominated feature importance in both datasets, confirming consistent biomarker identification across pediatric and adult populations. These findings demonstrate that calibrated gradient-boosted ensembles using interpretable EEG features achieve clinically safe seizure detection with cross-dataset generalizability. Cross-dataset external validation: The framework generalized to the Siena Scalp EEG Database, an independent adult focal epilepsy cohort, achieving 95% event-level sensitivity (Extra Trees) and AUC of 0.86 (Random Forest). Temporal lobe features (T3, T4, and T6 skewness and theta-band power) dominated SHAP rankings in both datasets, demonstrating that the identified biomarkers generalize across pediatric and adult populations with different recording equipment and clinical protocols. [ 19 ] examined dataset limitations, noting the dominance of the CHB–MIT corpus and the gap between benchmark performance and real-world applicability. These reviews highlight the need for interpretable, patient-generalizable This system can continuously monitor EEG for 24 h or longer without spurious alarms, a prerequisite for deployment in intensive care and ambulatory environments. This supports Hypothesis H 3 : calibrated gradient-boosted ensembles deliver clinically safe, patient-independent seizure detection. 4.8. External Validation on Siena Dataset To evaluate cross-dataset generalization, the trained ensemble classifiers were applied to the Siena Scalp EEG Database using identical preprocessing and feature extraction pipelines. Table 15 summarizes model performance across 12 adult patients with focal epilepsy under LOSO evaluation. Extra Trees achieved the highest event-level sensitivity, detecting 35 of 37 seizure events (95%), while Random Forest provided the best balance between discrimination (AUC = 0.86) and specificity (90%). The gradient-boosted models (XGBoost, LightGBM) maintained high specificity (>97%) with more conservative detection thresholds. These results demonstrate that the ensemble framework generalizes effectively to an independent dataset with different patient demographics (adult vs. pediatric), recording equipment, and clinical protocols. Table 16 compares performance between within-dataset (CHB–MIT) and external (Siena) evaluation. Seizure morphology, propagation patterns, and background EEG characteristics differ systematically between pediatric and adult populations. 3. Recording equipment: Different amplifier systems (CHB–MIT: unknown clinical system; Siena: BI 9800 and Galileo NT) introduce site-specific signal characteristics, including noise profiles, filtering responses, and amplitude calibration differences. 4. Power-line interference: CHB–MIT recordings required 60 Hz notch filtering (North American power grid), whereas Siena recordings required 50 Hz filtering (European power grid). 100%) reflects a domain shift between pediatric and adult populations, different recording equipment, and varying seizure morphologies. Site-specific threshold calibration would likely improve specificity for clinical deployment. Future work will evaluate the framework on additional multi-center datasets spanning diverse epilepsy subtypes to establish broader generalizability, as suggested by prior studies [ 4 ]. Temporal dynamics. The analysis operated at the segment level (5 s windows) with limited temporal context. While temporal smoothing improved event-level detection, it did not explicitly model seizure evolution. Using standardized preprocessing and feature engineering on the CHB–MIT pediatric corpus, the system achieved near-perfect discrimination (AUC ≥ 0.99) with zero false alarms under strict clinical safety constraints. XGBoost provided the best balance of performance and reliability, combining a sensitivity above 90% with perfect specificity, supported by excellent probability calibration. These results demonstrate that multi-domain, physiologically grounded EEG features combined with calibrated gradient-boosted ensembles attain reliable, low-false-alarm performance under patient-independent evaluation. External validation on the Siena Scalp EEG Database confirmed cross-dataset generalization, with 95% event-level sensitivity and consistent temporal lobe biomarker identification across pediatric and adult populations. Future work incorporating adaptive calibration, temporal modeling, and multi-center validation will be essential for translation into continuous bedside and ambulatory seizure surveillance. Figure 1 Representative scalp EEG segments from the CHB–MIT pediatric database. ( a ) Ten-second interictal segment from patient chb01, channel FP1-F7, showing normal background activity with preserved alpha rhythm.