From bed to bench and back again: Challenges facing deployment of intracranial pressure data analysis in clinical environments.

Introduction: Numerous complex physiological models derived from intracranial pressure (ICP) monitoring data have been developed. More recently, techniques such as machine learning are being used to develop increasingly sophisticated models to aid in clinical decision-making tasks such as diagnosis and prediction. Whilst their potential clinical impact may be significant, few models based on ICP data are routinely available at a patient's bedside. Further, the ability to refine models using ongoing patient data collection is rare. In this paper we identify and discuss the challenges faced when converting insight from ICP data analysis into deployable tools at the patient bedside. Research question: To provide an overview of challenges facing implementation of sophisticated ICP models and analyses at the patient bedside. Material and methods: A narrative review of the barriers facing implementation of sophisticated ICP models and analyses at the patient bedside in a neurocritical care unit combined with a descriptive case study (the CHART-ADAPT project) on the topic. Results: Key barriers found were technical, analytical, and integrity related. Examples included: lack of interoperability of medical devices for data collection and/or model deployment; inadequate infrastructure, hindering analysis of large volumes of high frequency patient data; a lack of clinical confidence in a model; and ethical, trust, security and patient confidentiality considerations governing the secondary use of patient data. Discussion and conclusion: To realise the benefits of ICP data analysis, the results need to be promptly delivered and meaningfully communicated. Multiple barriers to implementation remain and solutions which address real-world challenges are required.

A clinical evaluation of variation in paediatric intracranial pressure waveforms.

Introduction: Intracranial pressure (ICP) monitoring is commonly used in investigating the aetiology of chronic paediatric neurological conditions. A series of high-amplitude spikes has been observed in overnight ICP recordings of some children, many of whom have hydrocephalus or craniosynostosis. Research question: This clinical evaluation aimed to define the spike pattern, describe the patient group in which it is most likely to occur, and conduct high-resolution waveform analysis. Material and methods: ICP waveforms from 40 patients aged 0-5 years (inclusive), recorded between 2017 and 2021 at the Royal Hospital for Children Glasgow, were retrospectively analysed. The pattern was defined through visual inspection of regions of interest by two reviewers. Patients were stratified using demographic and clinical data. R software was used to perform regression and high-resolution waveform analyses. Results: The spike pattern was defined as the presence of 2 consecutive spikes with an amplitude of at least 8 mmHg, with a gap of at least 30 min between spikes. In the adjusted Poisson regression, age was significantly associated with the number of spikes (IRR 0.8, 95% CI 0.70 to 0.92, p-value 0.001). Discussion and conclusion: Younger age was significantly associated with an increased number of spikes in this cohort. Investigation of clinical consequences of the spikes is warranted.

International e-Delphi survey to define best practice in the reporting of intracranial pressure monitoring recording data.

Introduction: Intracranial pressure (ICP) monitoring is a very commonly performed neurosurgical procedure but there is a wide variation in how it is reported, hindering analysis of it. The current study sought to generate consensus on the reporting of ICP monitoring recording data. Research question: "What should be included in an ICP monitoring report?" Material and methods: The exercise was completed via a modified eDelphi survey. An expert panel discussion was held from which themes were identified and used to produce a code to annotate the transcript of the discussion. Statements were generated for a further two rounds of electronic questionnaires distributed via the REDcap platform. A Likert scale was used to grade agreement with each statement in the survey. A statement was accepted if more than 70% agreement was achieved between respondents. Data was collated using Microsoft Excel and analysed using R. Results: 149 relevant statements were identified from the transcript and categorised into recording parameters, waveform characteristics or reporting. A total of 22 statements were generated for the first round of the survey which was answered by 39 respondents. Following the electronic round of surveys consensus was achieved for all but one statement regarding the acceptability of automating ICP reporting. This was put forward to a second round after which 79% agreement was reached. Discussion and conclusion: The themes and statements from this eDelphi can be used as a framework to allow the standardisation of the reporting of intracranial pressure monitoring data.

Feasibility of forecasting future critical care bed availability using bed management data.

OBJECTIVES: This project aims to determine the feasibility of predicting future critical care bed availability using data-driven computational forecast modelling and routinely collected hospital bed management data. METHODS: In this proof-of-concept, single-centre data informatics feasibility study, regression-based and classification data science techniques were applied retrospectively to prospectively collect routine hospital-wide bed management data to forecast critical care bed capacity. The availability of at least one critical care bed was forecasted using a forecast horizon of 1, 7 and 14 days in advance. RESULTS: We demonstrated for the first time the feasibility of forecasting critical care bed capacity without requiring detailed patient-level data using only routinely collected hospital bed management data and interpretable models. Predictive performance for bed availability 1 day in the future was better than 14 days (mean absolute error 1.33 vs 1.61 and area under the curve 0.78 vs 0.73, respectively). By analysing feature importance, we demonstrated that the models relied mainly on critical care and temporal data rather than data from other wards in the hospital. DISCUSSION: Our data-driven forecasting tool only required hospital bed management data to forecast critical care bed availability. This novel approach means no patient-sensitive data are required in the modelling and warrants further work to refine this approach in future bed availability forecast in other hospital wards. CONCLUSIONS: Data-driven critical care bed availability prediction was possible. Further investigations into its utility in multicentre critical care settings or in other clinical settings are warranted.

Exploration of the relationship between partial pressure of brain tissue oxygen and intracranial pressure.

Introduction: Partial pressure of brain tissue oxygen (PbtO2) has been shown to be a safe an effective monitoring modality to compliment intracranial pressure (ICP) monitoring. It is related to metabolic activity, disease severity and mortality. Research question: Understanding the complex relationship between PbtO2 and ICP for patients with traumatic brain injury will enable better clinical decision making beyond simple threshold treatment strategies. Material and methods: Patients with PbtO2 monitoring were identified from the BrainIT database, a multi-centre dataset, containing minute by minute PbtO2 and ICP readings. Missing data was imputed and a multi-level log-normal regression model with a compound symmetry correlation structure was built. This accounted for any increased correlation due to the repeated measurements. The model was adjusted for mean arterial pressure and the partial pressure of carbon dioxide. Non-linearity was assessed using analysis of deviance and trends using expected marginal means. Results: 11 subjects with over 82,000 readings were included. They had a median age of 38 (IQR: 37-47), 73% were male, a median length of stay of 11.8 (IQR: 6.6-19.7) days and a median extended Glasgow outcome scale of 7.00 (IQR: 5-8).There is a statistically significant (p < 0.001) non-linear effect of ICP on PbtO2. With an overall increase in PbtO2 of 5.2% (95% CI 4%-6.4%, p < 0.001) for a 10 mmHg increase in ICP below 22 mmHg and a decrease of 5.5% (95% CI 2.7%-8.3%, p=<0.001) in PbtO2 for a 10 mmHg increase in ICP above 22 mmHg. As well as a decrease of 40.9% (95% CI 2.3%-64.3%, p = 0.040) in PbtO2 per day in the intensive care unit. Discussion and conclusion: This model demonstrates that there is a significant non-linear relationship between ICP and PbtO2, however, this is a small heterogeneous cohort and further validation will be required.

The impact of electronic versus paper-based data capture on data collection logistics and on missing scores in thyroid cancer patients.

PURPOSE: The purpose of this study was to investigate the impact of the type of data capture on the time and help needed for collecting patient-reported outcomes as well as on the proportion of missing scores. METHODS: In a multinational prospective study, thyroid cancer patients from 17 countries completed a validated questionnaire measuring quality of life. Electronic data capture was compared to the paper-based approach using multivariate logistic regression. RESULTS: A total of 437 patients were included, of whom 13% used electronic data capture. The relation between data capture and time needed was modified by the emotional functioning of the patients. Those with clinical impairments in that respect needed more time to complete the questionnaire when they used electronic data capture compared to paper and pencil (ORadj 24.0; p = 0.006). This was not the case when patients had sub-threshold emotional problems (ORadj 1.9; p = 0.48). The odds of having the researcher reading the questions out (instead of the patient doing this themselves) (ORadj 0.1; p = 0.01) and of needing any help (ORadj 0.1; p = 0.01) were lower when electronic data capture was used. The proportion of missing scores was equivalent in both groups (ORadj 0.4, p = 0.42). CONCLUSIONS: The advantages of electronic data capture, such as real-time assessment and fewer data entry errors, may come at the price of more time required for data collection when the patients have mental health problems. As this is not uncommon in thyroid cancer, researchers need to choose the type of data capture wisely for their particular research question.

International Phase IV Field Study for the Reliability and Validity of the European Organisation for Research and Treatment of Cancer Thyroid Cancer Module EORTC QLQ-THY34.

Purpose: The aim of this study was to validate the new European Organisation for Research and Treatment of Cancer Quality of Life Thyroid Cancer Module (EORTC QLQ-THY34). Methods: We enrolled 437 thyroid cancer patients from 17 countries. One group (n = 303), undergoing treatment or best supportive care, completed the questionnaires at three time points (before therapy [t1], 6 weeks later [t2], and 6 months after t2 [t3]). A second group (survivors ≥2 years after diagnosis, n = 134) completed it at a random baseline time point and a second time 1 week later. We determined internal consistency (using Cronbach's alpha), the scale structure (with confirmatory factor analysis), and discriminant validity (using known-group comparisons). Group 1 data were used to assess responsiveness and group 2 data to determine test-retest reliability using intra-class correlations (ICC). Results: All 34 items fulfilled the criteria to be kept in the questionnaire. Cronbach's alpha was >0.70 in 8 of the 9 multi-item scales. All standardized factor loadings exceeded 0.40, confirming the proposed scale structure. The ICC was >0.70 in all scales expressing good test-retest reliability. Differences in scale scores between patients with different histology were >5 points in all scales. In all but one of the pre-specified scales (Dry Mouth), changes over time were ≥|4| points between at least two time points. Conclusion: The EORTC QLQ-THY34 with its 9 multi-item and 8 single-item scales is a reliable and valid tool to measure quality of life in thyroid cancer patients and can be used in future trials and studies.

Decompressive craniectomy as a second/third-tier intervention in traumatic brain injury: A multicenter observational study.

OBJECTIVES: RESCUEicp studied decompressive craniectomy (DC) applied as third-tier option in severe traumatic brain injury (TBI) patients in a randomized controlled setting and demonstrated a decrease in mortality with similar rates of favorable outcome in the DC group compared to the medical management group. In many centers, DC is being used in combination with other second/third-tier therapies. The aim of the present study is to investigate outcomes from DC in a prospective non-RCT context. METHODS: This is a prospective observational study of 2 patient cohorts: one from the University Hospitals Leuven (2008-2016) and one from the Brain-IT study, a European multicenter database (2003-2005). In thirty-seven patients with refractory elevated intracranial pressure who underwent DC as a second/third-tier intervention, patient, injury and management variables including physiological monitoring data and administration of thiopental were analysed, as well as Extended Glasgow Outcome score (GOSE) at 6 months. RESULTS: In the current cohorts, patients were older than in the surgical RESCUEicp cohort (mean 39.6 vs. 32.3; p < 0.001), had higher Glasgow Motor Score on admission (GMS < 3 in 24.3% vs. 53.0%; p = 0.003) and 37.8% received thiopental (vs. 9.4%; p < 0.001). Other variables were not significantly different. GOSE distribution was: death 24.3%; vegetative 2.7%; lower severe disability 10.8%; upper severe disability 13.5%; lower moderate disability 5.4%; upper moderate disability 2.7%, lower good recovery 35.1%; and upper good recovery 5.4%. The outcome was unfavorable in 51.4% and favorable in 48.6%, as opposed to 72.6% and 27.4% respectively in RESCUEicp (p = 0.02). CONCLUSION: Outcomes in DC patients from two prospective cohorts reflecting everyday practice were better than in RESCUEicp surgical patients. Mortality was similar, but fewer patients remained vegetative or severely disabled and more patients had a good recovery. Although patients were older and injury severity was lower, a potential partial explanation may be in the pragmatic use of DC in combination with other second/third-tier therapies in real-life cohorts. The findings underscore that DC maintains an important role in managing severe TBI.

Patient preference for commonly-used, head and neck cancer-specific quality of life questionnaires in the follow-up setting (Determin): A multi-centre randomised controlled trial and mixed methods study.

BACKGROUND: Quality of life (QoL) assessment forms an integral part of modern cancer care and research. The aim of this study is to determine patients' preferences and willingness to complete commonly used head-and-neck cancer (HNC) QoL questionnaires (QLQs) in routine follow-up clinics. METHODS: This is a randomised control trial of 583 subjects from 17 centres during follow-up after treatment for oral, oropharyngeal or laryngeal cancer. Subjects completed three structured validated questionnaires: EORTC QLQ-HN35; FACT-HN and UW-QOL, and an unstructured patient-generated list. The order of questionnaire presentation was randomised, and subjects were stratified by disease site and stage. Patients self-rated the questionnaires they found most helpful to communicate their health concerns to their clinicians. RESULTS: Of the 558 respondents, 82% (457) found QLQs useful to communicate their health concerns to their clinician (OR = 15.76; 95% CI 10.83-22.94). Patients preferred the structured disease-specific instruments (OR 8.79; 95% CI 5.99-12.91), while the open list was the most disliked (OR = 4.25; 95% CI 3.04-5.94). There was no difference in preference by treatment modality. More women preferred the FACT-HN (OR = 3.01, 95% CI 1.05-8.62), and patients under 70 preferred EORTC QLQ-HN35 (OR = 3.14, 95% CI 1.3-7.59). However, only 55% of patients expressed preference to complete questionnaires routinely at the clinic. CONCLUSIONS: Most patients found QLQs helpful during their follow-up and 55% supported routine questionnaires in follow-up clinics. Males and people over 70 years old were the least willing to complete the routine questionnaires and preferred shorter questionnaires (e.g., UW-QOL). Women preferred FACT-HN, and younger patients preferred EORTC QLQ-HN35. Reasons for the reluctance to complete questionnaires require elucidation.

Application of synthetic data in the training of artificial intelligence for automated quality assurance in magnetic resonance imaging.

BACKGROUND: Magnetic resonance imaging scanner faults can be missed during routine quality assurance (QA) if they are subtle, intermittent, or the test being performed is insensitive to the type of fault. Coil element malfunction is a common fault within MRI scanners, which may go undetected for quite some time. Consequently, this may lead to poor image quality and the potential for misdiagnoses. PURPOSE: Daily QA typically consists of an automated signal to noise ratio test and in some instances this test is insensitive to coil element malfunction. Instead of relying on daily QA testing, it was proposed to utilize patient images in conjunction with a trained neural network to detect coil element malfunction, even when it presents as a very subtle defect. The advantage to using patient images over phantom testing is real-time monitoring can be achieved. This allows clinical staff to focus more on patient throughput without being burdened by daily testing. METHODS: A neural network was trained using simulated coil failure in 3958 abdominal or pelvic images from 497 patients. The accuracy of the trained network was then tested on an unseen dataset of 109 images from which 44 patients which had coil element malfunction present. Five MRI radiographers were shown 249 images with and without real coil malfunction to assess their accuracy compared to the neural network in identifying the scanner fault. RESULTS: The neural network achieved an accuracy of 91.74% in identifying coil element malfunction in the unseen data. Radiographers tasked with identifying coil element malfunction had an average accuracy of 59.99%. In the same test case, the neural network outperformed all radiographers with an accuracy of 91.56%. CONCLUSION: This work demonstrates that neural networks trained with artificial data can successfully identify MRI scanner coil element malfunction in clinical images. The method provided better accuracy than MRI radiographers (technologists) at identifying coil element malfunction and highlights the potential utility of AI methods as an alternative to support traditional QA. Further, our methodology of training neural networks with simulated data could potentially identify other faults, allowing centers to produce robust fault detection systems with minimal data.

Serum antibodies to periodontal pathogens prior to rheumatoid arthritis diagnosis: A case-control study.

OBJECTIVES: 1) To quantify the association between anti-Porphyromonas gingivalis serum antibody concentrations and the risk of developing rheumatoid arthritis (RA), and 2) to quantify the associations among RA cases between anti-P. gingivalis serum antibody concentrations and RA-specific autoantibodies. Additional anti-bacterial antibodies evaluated included anti-Fusobacterium nucleatum and anti-Prevotella intermedia. METHODS: Serum samples were acquired pre- and post- RA diagnosis from the U.S. Department of Defense Serum Repository (n = 214 cases, 210 matched controls). Using separate mixed-models, the timing of elevations of anti-P. gingivalis, anti-P. intermedia, and anti-F. nucleatum antibody concentrations relative to RA diagnosis were compared in RA cases versus controls. Associations were determined between serum anti-CCP2, ACPA fine specificities (vimentin, histone, and alpha-enolase), and IgA, IgG, and IgM RF in pre-RA diagnosis samples and anti-bacterial antibodies using mixed-effects linear regression models. RESULTS: No compelling evidence of case-control divergence in serum anti-P. gingivalis, anti-F. nucleatum, and anti-P. intermedia was observed. Among RA cases, including all pre-diagnosis serum samples, anti-P. intermedia was significantly positively associated with anti-CCP2, ACPA fine specificities targeting vimentin, histone, alpha-enolase, and IgA RF (p<0.001), IgG RF (p = 0.049), and IgM RF (p = 0.004), while anti-P. gingivalis and anti-F. nucleatum were not. CONCLUSIONS: No longitudinal elevations of anti-bacterial serum antibody concentrations were observed in RA patients prior to RA diagnosis compared to controls. However, anti-P. intermedia displayed significant associations with RA autoantibody concentrations prior to RA diagnosis, suggesting a potential role of this organism in progression towards clinically-detectable RA.

Demystifying the Black Box: The Importance of Interpretability of Predictive Models in Neurocritical Care.

Neurocritical care patients are a complex patient population, and to aid clinical decision-making, many models and scoring systems have previously been developed. More recently, techniques from the field of machine learning have been applied to neurocritical care patient data to develop models with high levels of predictive accuracy. However, although these recent models appear clinically promising, their interpretability has often not been considered and they tend to be black box models, making it extremely difficult to understand how the model came to its conclusion. Interpretable machine learning methods have the potential to provide the means to overcome some of these issues but are largely unexplored within the neurocritical care domain. This article examines existing models used in neurocritical care from the perspective of interpretability. Further, the use of interpretable machine learning will be explored, in particular the potential benefits and drawbacks that the techniques may have when applied to neurocritical care data. Finding a solution to the lack of model explanation, transparency, and accountability is important because these issues have the potential to contribute to model trust and clinical acceptance, and, increasingly, regulation is stipulating a right to explanation for decisions made by models and algorithms. To ensure that the prospective gains from sophisticated predictive models to neurocritical care provision can be realized, it is imperative that interpretability of these models is fully considered.

Low iodine diet advice and differentiated thyroid cancer treatment: A historic exploration in three UK centres.

BACKGROUND AND AIMS: Patients with differentiated thyroid cancer are often advised to follow a low iodine diet (LID) one to two weeks before radioiodine remnant ablation (RRA). We describe treatment practices and ablation success rates in centres (C1, C2, C3) in the UK with different approaches to LID advice. METHODS: Historic cohort of patients with differentiated thyroid cancer treated with RRA in 2015/16 in C1 (n = 50, 1-week LID), C2 (n = 59, 2-week LID) and C3 (n = 108, no LID advice). Response to RRA was stratified as excellent, indeterminate, or incomplete by the adapted American Thyroid Association Dynamic Risk Stratification Score. RESULTS: There was little difference in age, sex and staging between centres, but the percentage receiving 1.1 GBq vs higher administered activities differed (C1:22%, C2:44%, C3:15%, p < 0.001). Excellent response was recorded for: C1:48%, C2:36%, C3:49% (p = 0.61). Differences in RRA preparation and outcome assessment at C3 precluded comparison across all centres. Adjusted odds ratio for excellent response at C2 vs C1 was 0.57 (95%CI: 0.25,1.32), p = 0.19. CONCLUSIONS: There was no evidence that advising a LID for 2-weeks before RRA improves outcomes compared to 1-week. For definitive recommendations on LIDs prior to RRA, a prospective multi-centre study with a more homogenous approach to patient management or, randomised controlled trial, is needed.

Improving the Estimation of Subgroup Effects for Clinical Trial Participants with Multimorbidity by Incorporating Drug Class-Level Information in Bayesian Hierarchical Models: A Simulation Study.

BACKGROUND: There is limited guidance for using common drug therapies in the context of multimorbidity. In part, this is because their effectiveness for patients with specific comorbidities cannot easily be established using subgroup analyses in clinical trials. Here, we use simulations to explore the feasibility and implications of concurrently estimating effects of related drug treatments in patients with multimorbidity by partially pooling subgroup efficacy estimates across trials. METHODS: We performed simulations based on the characteristics of 161 real clinical trials of noninsulin glucose-lowering drugs for diabetes, estimating subgroup effects for patients with a hypothetical comorbidity across related trials in different scenarios using Bayesian hierarchical generalized linear models. We structured models according to an established ontology-the World Health Organization Anatomic Chemical Therapeutic Classifications-allowing us to nest all trials within drugs and all drugs within anatomic chemical therapeutic classes, with effects partially pooled at each level of the hierarchy. In a range of scenarios, we compared the performance of this model to random effects meta-analyses of all drugs individually. RESULTS: Hierarchical, ontology-based Bayesian models were unbiased and accurately recovered simulated comorbidity-drug interactions. Compared with single-drug meta-analyses, they offered a relative increase in precision of up to 250% in some scenarios because of information sharing across the hierarchy. Because of the relative precision of the approaches, a large proportion of small subgroup effects was detectable only using the hierarchical model. CONCLUSIONS: By assuming that similar drugs may have similar subgroup effects, Bayesian hierarchical models based on structures defined by existing ontologies can be used to improve the precision of treatment efficacy estimates in patients with multimorbidity, with potential implications for clinical decision making.

Combinations of Anticyclic Citrullinated Protein Antibody, Rheumatoid Factor, and Serum Calprotectin Positivity Are Associated With the Diagnosis of Rheumatoid Arthritis Within 3 Years.

OBJECTIVE: To evaluate the prevalence of elevations of anti-cyclic citrullinated peptide-3 (anti-CCP3) antibody, rheumatoid factor IgM (RF-IgM) and serum calprotectin (sCP) in pre-rheumatoid arthritis (RA) as well as the diagnostic accuracies of these biomarkers for the timing of diagnosis of future RA. METHODS: A total of 215 RA cases, each with approximately three pre-RA diagnoses and one post-RA diagnosis serum sample, and controls were identified from the Department of Defense Serum Repository. All case samples and a single sample from each control subject were tested for anti-CCP3 (IgG), RF-IgM, and sCP. The diagnostic accuracies of biomarkers for future RA were evaluated. RESULTS: Anti-CCP3, RF-IgM, and sCP were elevated in pre-RA, with anti-CCP3 and sCP significantly elevated compared with RF-IgM at the earliest time points. Within the cases, the combination of anti-CCP3 and RF-IgM positivity had a positive predictive value (PPV) of 35.6% for a diagnosis of RA in 3 years or less, which is significantly higher than the PPV of 18.7% for anti-CCP3 positivity alone (P < 0.001). A combination of anti-CCP3, RF-IgM, and sCP had the highest PPV (53.0%) for a diagnosis of RA in 3 years or less; however, this was not significantly higher than the PPV for anti-CCP3 and RF-IgM positivity (P = 0.248). CONCLUSION: Anti-CCP3, RF-IgM, and sCP are elevated in pre-RA; furthermore, combinations of elevations of these biomarkers are more commonly seen in the period of less than or equal to 3 years to diagnosis. This may be considered in creating inclusion criteria in prevention trials in RA. In addition, the biologic relationships of these biomarkers in pre-RA need exploration.

Use of Direct Intracranial Pressure and Brain Tissue Oxygen Monitoring in Perioperative Management of Patients with Moyamoya Disease.

Intracranial pressure monitoring and brain tissue oxygen monitoring are commonly used in head injury for goal-directed therapies, but there may be more indications for its use. Moyamoya disease involves progressive stenosis of the arterial circulation and formation of collateral vessels that are at risk of hemorrhage. The risk of ischemic events during revascularization surgery and postoperatively is high. Impaired cerebral autoregulation may be one of the factors that are implicated. We present our experience with monitoring of cerebral oxygenation and autoregulation in the pathological hemisphere during the perioperative period in four patients with moyamoya disease.

Influence of Patient Demographics on Optimal Cerebral Perfusion Pressure Following Traumatic Brain Injury.

The relationship between optimal cerebral perfusion pressure (CPPopt) and patient characteristics has yet to be defined but could have significant implications for future guidelines recommending cerebral perfusion pressure (CPP) targets.Data from 36 traumatic brain injured patients admitted to neurological intensive care were analysed retrospectively. Linear mixed effects (LME) analysis was performed using an unadjusted-adjusted approach.Clinical characteristics with p < 0.10 were included in the adjusted model. A second adjusted model which included all variables of interest was created. Model fit was assessed using the root-mean-square error (RMSE).The adjusted model included time from initiation of intracranial pressure (ICP) monitoring (estimate = 0.00292, p < 0.001), age (estimate = -0.211, p = 0.0750) and the presence of diffuse axonal injury (DAI) (estimate = -35.5, p < 0.001). The RMSE of this model was 8.11 mmHg. The RMSE of the model containing all variables was 8.09 mmHg.Time, age and the presence of DAI may be important predictors of CPPopt. The models were too inaccurate at predicting CPPopt for employment in clinical practice but warrant further investigation. CPPopt is a dynamic measurement influenced by many factors, supporting the utility of investigating the feasibility of CPPopt-guided therapy.

Evaluation of Software for Automated Measurement of Adherence to ICP-Monitoring Threshold Guideline.

Challenges inherent in clinical guideline development include a long time lag between the key results and incorporation into best practice and the qualitative nature of adherence measurement, meaning it will have no directly measurable impact. To address these issues, a framework has been developed to automatically measure adherence by clinicians in neurological intensive care units to the Brain Trauma Foundation's intracranial pressure (ICP)-monitoring guidelines for severe traumatic brain injury (TBI).The framework processes physiological and treatment data taken from the bedside, standardises the data as a set of process models, then compares these models against similar process models constructed from published guidelines. A similarity metric (i.e. adherence measure) between the two models is calculated, composed of duration and scale of non-adherence.In a pilot clinical validation test, the framework was applied to physiological/treatment data from three TBI patients exhibiting ICP secondary insults at a local neuro-centre where clinical experts coded key clinical interventions/decisions about patient management.The framework identified non-adherence with respect to drug administration in one patient, with a spike in non-adherence due to an inappropriately high dosage; a second patient showed a high severity of guideline non-adherence; and a third patient showed non-adherence due to a low number of associated events and treatment annotations.

Time Series Analysis and Prediction of Intracranial Pressure Using Time-Varying Dynamic Linear Models.

Intracranial pressure (ICP) monitoring is a key clinical tool in the assessment and treatment of patients in a neuro-intensive care unit (neuro-ICU). As such, a deeper understanding of how an individual patient's ICP can be influenced by therapeutic interventions could improve clinical decision-making. A pilot application of a time-varying dynamic linear model was conducted using the BrainIT dataset, a multi-centre European dataset containing temporaneous treatment and vital-sign recordings. The study included 106 patients with a minimum of 27 h of ICP monitoring. The model was trained on the first 24 h of each patient's ICU stay, and then the next 2 h of ICP was forecast. The algorithm enabled switching between three interventional states: analgesia, osmotic therapy and paralysis, with the inclusion of arterial blood pressure, age and gender as exogenous regressors. The overall median absolute error was 2.98 (2.41-5.24) mmHg calculated using all 106 2-h forecasts. This is a novel technique which shows some promise for forecasting ICP with an adequate accuracy of approximately 3 mmHg. Further optimisation is required for the algorithm to become a usable clinical tool.