Clusters of medical specialties around patients with multimorbidity - employing fuzzy c-means clustering to explore multidisciplinary collaboration.

BACKGROUND: Hospital care organization, structured around medical specialties and focused on the separate treatment of individual organ systems, is challenged by the increasing prevalence of multimorbidity. To support the hospitals' realization of multidisciplinary care, we hypothesized that using machine learning on clinical data helps to identify groups of medical specialties who are simultaneously involved in hospital care for patients with multimorbidity. METHODS: We conducted a cross-sectional study of patients in a Dutch general hospital and used a fuzzy c-means clustering algorithm for the analysis. We explored the patients' membership degrees in each cluster to identify subgroups of medical specialties that provide care to the same patients with multimorbidity. We used retrospectively collected electronic health record data from 2017. We extracted data from 22,133 patients aged ≥18 years who had received outpatient clinical care for two or more chronic and/ or oncological diagnoses. RESULTS: We found six clusters of medical specialties and identified 22 subgroups. The clusters were labeled based on the specialties that most characterized them: 1. dermatology/ plastic surgery, 2. six specialties (gynecology/ rheumatology/ orthopedic surgery/ urology/ gastroenterology/ otorhinolaryngology), 3. pulmonology, 4. internal medicine/ cardiology/ geriatrics, 5. neurology/ physiatry (rehabilitation)/ anesthesiology, and 6. internal medicine. Most patients had a full or dominant membership to one of these clusters of medical specialties (11 subgroups), whereas fewer patients had a membership to two clusters. The prevalence of specific diagnosis groups, patient characteristics, and healthcare utilization differed between subgroups. CONCLUSION: Our study shows that clusters and subgroups of medical specialties simultaneously involved in hospital care for patients with multimorbidity can be identified with fuzzy c-means cluster analysis using clinical data. Clusters and subgroups differed regarding the involved medical specialties, diagnoses, patient characteristics, and healthcare utilization. With this strategy, hospitals and medical specialists can further analyze which subgroups are target populations that might benefit from improved multidisciplinary collaboration.

The Dynamics Underlying the Rise of Star Performers.

Across different domains, there are 'star performers' who are able to generate disproportionate levels of performance output. To date, little is known about the model principles underlying the rise of star performers. Here, we propose that star performers' abilities develop according to a multi-dimensional, multiplicative and dynamical process. Based on existing literature, we defined a dynamic network model, including different parameters functioning as enhancers or inhibitors of star performance. The enhancers were multiplicity of productivity, monopolistic productivity, job autonomy, and job complexity, whereas productivity ceiling was an inhibitor. These enhancers and inhibitors were expected to influence the tail-heaviness of the performance distribution. We therefore simulated several samples of performers, thereby including the assumed enhancers and inhibitors in the dynamic networks and compared their tail-heaviness. Results showed that the dynamic network model resulted in heavier and lighter tail distributions, when including the enhancer- and inhibitor-parameters, respectively. Together, these results provide novel insights into the dynamical principles that give rise to star performers in the population.

The value of artificial intelligence for the treatment of mechanically ventilated intensive care unit patients: An early health technology assessment.

PURPOSE: The health and economic consequences of artificial intelligence (AI) systems for mechanically ventilated intensive care unit patients often remain unstudied. Early health technology assessments (HTA) can examine the potential impact of AI systems by using available data and simulations. Therefore, we developed a generic health-economic model suitable for early HTA of AI systems for mechanically ventilated patients. MATERIALS AND METHODS: Our generic health-economic model simulates mechanically ventilated patients from their hospitalisation until their death. The model simulates two scenarios, care as usual and care with the AI system, and compares these scenarios to estimate their cost-effectiveness. RESULTS: The generic health-economic model we developed is suitable for estimating the cost-effectiveness of various AI systems. By varying input parameters and assumptions, the model can examine the cost-effectiveness of AI systems across a wide range of different clinical settings. CONCLUSIONS: Using the proposed generic health-economic model, investors and innovators can easily assess whether implementing a certain AI system is likely to be cost-effective before an exact clinical impact is determined. The results of the early HTA can aid investors and innovators in deployment of AI systems by supporting development decisions, informing value-based pricing, clinical trial design, and selection of target patient groups.

Mechanical ventilation as a major driver of COVID-19 hospitalization costs: a costing study in a German setting.

BACKGROUND: While COVID-19 hospitalization costs are essential for policymakers to make informed health care resource decisions, little is known about these costs in western Europe. The aim of the current study is to analyze these costs for a German setting, track the development of these costs over time and analyze the daily costs. METHODS: Administrative costing data was analyzed for 598 non-Intensive Care Unit (ICU) patients and 510 ICU patients diagnosed with COVID-19 at the Frankfurt University hospital. Descriptive statistics of total per patient hospitalization costs were obtained and assessed over time. Propensity scores were estimated for length of stay (LOS) at the general ward and mechanical ventilation (MV) duration, using covariate balancing propensity score for continuous treatment. Costs for each additional day in the general ward and each additional day in the ICU with and without MV were estimated by regressing the total hospitalization costs on the LOS and the presence or absence of several treatments using generalized linear models, while controlling for patient characteristics, comorbidities, and complications. RESULTS: Median total per patient hospitalization costs were €3,010 (Q1 - Q3: €2,224-€5,273), €5,887 (Q1 - Q3: €3,054-€10,879) and €21,536 (Q1 - Q3: €7,504-€43,480), respectively, for non-ICU patients, non-MV and MV ICU patients. Total per patient hospitalization costs for non-ICU patients showed a slight increase over time, while total per patient hospitalization costs for ICU patients decreased over time. Each additional day in the general ward for non-ICU COVID-19 patients costed €463.66 (SE: 15.89). Costs for each additional day in the general ward and ICU without and with mechanical ventilation for ICU patients were estimated at €414.20 (SE: 22.17), €927.45 (SE: 45.52) and €2,224.84 (SE: 70.24). CONCLUSIONS: This is, to our knowledge, the first study examining the costs of COVID-19 hospitalizations in Germany. Estimated costs were overall in agreement with costs found in literature for non-COVID-19 patients, except for higher estimated costs for mechanical ventilation. These estimated costs can potentially improve the precision of COVID-19 cost effectiveness studies in Germany and will thereby allow health care policymakers to provide better informed health care resource decisions in the future.

Identifying the need for infection-related consultations in intensive care patients using machine learning models.

Infection-related consultations on intensive care units (ICU) have a positive impact on quality of care and clinical outcome. However, timing of these consultations is essential and to date they are typically event-triggered and reactive. Here, we investigate a proactive approach to identify patients in need for infection-related consultations by machine learning models using routine electronic health records. Data was retrieved from a mixed ICU at a large academic tertiary care hospital including 9684 admissions. Infection-related consultations were predicted using logistic regression, random forest, gradient boosting machines, and long short-term memory neural networks (LSTM). Overall, 7.8% of admitted patients received an infection-related consultation. Time-sensitive modelling approaches performed better than static approaches. Using LSTM resulted in the prediction of infection-related consultations in the next clinical shift (up to eight hours in advance) with an area under the receiver operating curve (AUROC) of 0.921 and an area under the precision recall curve (AUPRC) of 0.541. The successful prediction of infection-related consultations for ICU patients was done without the use of classical triggers, such as (interim) microbiology reports. Predicting this key event can potentially streamline ICU and consultant workflows and improve care as well as outcome for critically ill patients with (suspected) infections.

CAPICE: a computational method for Consequence-Agnostic Pathogenicity Interpretation of Clinical Exome variations.

Exome sequencing is now mainstream in clinical practice. However, identification of pathogenic Mendelian variants remains time-consuming, in part, because the limited accuracy of current computational prediction methods requires manual classification by experts. Here we introduce CAPICE, a new machine-learning-based method for prioritizing pathogenic variants, including SNVs and short InDels. CAPICE outperforms the best general (CADD, GAVIN) and consequence-type-specific (REVEL, ClinPred) computational prediction methods, for both rare and ultra-rare variants. CAPICE is easily added to diagnostic pipelines as pre-computed score file or command-line software, or using online MOLGENIS web service with API. Download CAPICE for free and open-source (LGPLv3) at https://github.com/molgenis/capice .