Diagnosing Hemophagocytic Lymphohistiocytosis with Machine Learning: A Proof of Concept.

Hemophagocytic lymphohistiocytosis is a hyperinflammatory syndrome characterized by uncontrolled activation of immune cells and mediators. Two diagnostic tools are widely used in clinical practice: the HLH-2004 criteria and the Hscore. Despite their good diagnostic performance, these scores were constructed after a selection of variables based on expert consensus. We propose here a machine learning approach to build a classification model for HLH in a cohort of patients selected by glycosylated ferritin dosage in our tertiary center in Lyon, France. On a dataset of 207 adult patients with 26 variables, our model showed good overall diagnostic performances with a sensitivity of 71.4% and high specificity, and positive and negative predictive values which were 100%, 100%, and 96.9%, respectively. Although generalization is difficult on a selected population, this is the first study to date to provide a machine-learning model for HLH detection. Further studies will be required to improve the machine learning model performances with a large number of HLH cases and with appropriate controls.

Innovative Approach for a Typology of Treatment Sequences in Early Stage HER2 Positive Breast Cancer Patients Treated With Trastuzumab in the French National Hospital Database.

Background: Our objective was to describe the hospital-based systemic treatment sequences in early stage HER2+ breast cancer patients treated with trastuzumab in France in 2016. Methods: This retrospective observational study was based on the national hospital discharge database (PMSI). Patients hospitalized for breast cancer in 2016 and administration of trastuzumab between 6 months prior and 1 year after surgery were included. The following treatments were identified: (1) trastuzumab ± chemotherapy; (2) chemotherapy alone; (3) q3w trastuzumab weekly chemotherapy. Hospital admissions for cardiac events before and after the surgery were investigated. An unsupervised machine learning technic called TAK (Time-sequence Analysis through K-clustering) was used to identify and visualize typical systemic treatment sequences. Results: Overall, 3531 patients were included: 2619 adjuvant cohort patients (74.2%) and 912 neoadjuvant cohort patients (25.8%). The mean age was 56.4 years (±12.3), 99.7% patients were female. Treatment initiation occurred within 6 weeks of the surgery in 58% and 92% of patients, and trastuzumab treatment lasted 12 months (±1 month) in 75% and 66% of patients in the adjuvant and neoadjuvant cohorts, respectively. Nevertheless, 12% and 22% of patients were treated with trastuzumab for <11 months in the adjuvant and neoadjuvant cohorts, respectively. There was not one standard sequence of treatments per cohort, but 4 and 3 typical treatment sequences in the adjuvant and the neoadjuvant cohorts, respectively, plus 2 treatment sequences with an early treatment withdrawal. The frequency of patients with ⩾1 hospital stay with a cardiac event was higher among patients with an early treatment withdrawal. Conclusions: The treatment sequences of most patients were in line with the recommendations in force. The machine learning approach provided a telling visual display of the results, thereby allowing healthcare professionals, health authorities, patients, and care givers to see the whole picture of the hospital-administered drug strategies.

Machine Learning-Based Analysis of Treatment Sequences Typology in Advanced Non-Small-Cell Lung Cancer Long-Term Survivors Treated With Nivolumab.

PURPOSE: Immune checkpoint inhibitors substantially changed advanced non-small-cell lung cancer (aNSCLC) management and can lead to long-term survival. The aims of this study were (1) to use a machine learning method to establish a typology of treatment sequences on patients with aNSCLC who were alive 2 years after initiating a treatment with anti-programmed death-ligand 1 monoclonal antibody nivolumab and (2) to describe the patients' characteristics according to the typology of treatment sequences. MATERIALS AND METHODS: This retrospective observational study was based on data from the comprehensive French hospital discharge database for all patients with lung cancer with at least one line of platinum-based chemotherapy, starting nivolumab between January 1, 2015, and December 31, 2016, and alive 2 years after nivolumab treatment initiation. Patients were followed until December 31, 2018. A typology of most common treatment sequences was established using hierarchical clustering with time sequence analysis. RESULTS: Two thousand two hundred twelve study patients were, on average, 63.0 years old, 69.9% of them were men, and 61.9% had a nonsquamous cell carcinoma. During the 2 years after nivolumab treatment initiation, clusters of patients with four basic types of treatment sequences were identified: (1) almost continuous nivolumab treatment (44% of patients); (2) nivolumab most of the time followed by a treatment-free interval or a chemotherapy (15% of patients); and a short or medium nivolumab treatment, followed by (3) a long systemic treatment-free interval (17% of patients) or (4) a long chemotherapy (23% of patients). CONCLUSION: This machine learning approach enabled the identification of a typology of four representative treatment sequences observed in long-term survival. It was noted that most long-term survivors were treated with nivolumab for well over 1 year.

Costs and mortality associated with HIV: a machine learning analysis of the French national health insurance database.

BACKGROUND: The objective is to characterise the economic burden to the healthcare system of people living with HIV (PLWHIV) in France and to help decision makers in identifying risk factors associated with high-cost and high mortality profiles. DESIGN AND METHODS: The study is a retrospective analysis of PLWHIV identified in the French National Health Insurance database (SNDS). All PLWHIV present in the database in 2013 were identified.  All healthcare resource consumption from 2008 to 2015 inclusive was documented and costed (for 2013 to 2015) from the perspective of public health insurance. High-cost and high mortality patient profiles were identified by a machine learning algorithm. RESULTS: In 2013, 96,423 PLWHIV were identified in the SNDS database, including 3,373 incident cases. Overall, 3,224 PLWHIV died during the three-year follow-up period (mean annual mortality rate: 1.1%). The mean annual per capita cost incurred by PLWHIV was € 14,223, corresponding to a total management cost of HIV of € 1,370 million in 2013. The largest contribution came from the cost of antiretroviral medication (M€ 870; 63%) followed by hospitalisation (M€ 154; 11%). The costs incurred in the year preceding death were considerably higher. Four specific patient profiles were identified for under/over-expressing these costs, suggesting ways to reduce them. CONCLUSIONS: Even though current therapeutic regimens provide excellent virological control in most patients, PLWHIV have excess mortality. Other factors such as comorbidities, lifestyle factors and screening for cancer and cardiovascular disease, need to be targeted in order to lower the mortality and cost associated with HIV infection.

Factors Associated With Aggressiveness of End-of-Life Care for Lung Cancer Patients and Associated Costs of Care.

BACKGROUND: Results of previous studies demonstrated that high-intensity end-of-life (EOL) care improves neither cancer patients' survival nor quality of life. Our objective was to assess the incidence of and factors associated with aggressiveness of care during the last 30 days of life (DOL) of lung cancer (LC) patients and the impacts of aggressiveness of care in EOL-care costs. PATIENTS AND METHODS: Using French national hospital database, all patients with LC who died between January 1, 2010, and December 31, 2011, or between January 1, 2015, and January 31, 2016, were included. EOL-care aggressiveness was assessed using the following criteria: chemotherapy administered within the last 14 DOL; more than one hospitalization within the last 30 DOL; admission to the intensive care unit within the last 30 DOL; and palliative care initiated < 3 days before death. Expenditures were limited to direct costs, from a health care payer's perspective. RESULTS: Among 79,746 adult LC patients identified; 57% had at least one indicator of EOL-care aggressiveness (49% repeated hospitalizations, 12% intensive care unit admissions, 9% chemotherapy, 5% palliative care). It increased significantly between the 2 periods (56% vs. 58%, P < .001). Young age, male sex, shorter time since diagnosis, comorbidities, no malnutrition, type of care facility other than general hospital, social deprivation, and low-density population were independently associated with having one or more indicator of aggressive EOL care. The mean EOL cost was €8152 ± 5117 per patient, but the cost was significantly higher for patients with at least one EOL-care aggressiveness criterion (€9480 vs. €6376, P < .001). CONCLUSION: In France, a majority of LC patients had at least one criterion of aggressive EOL care that had a major economic impact on the health care system.

Automatic and Explainable Labeling of Medical Event Logs With Autoencoding.

Process mining is a suitable method for knowledge extraction from patient pathways. Structured in event logs, medical events are complex, often described using various medical codes. An efficient labeling of these events before applying process mining analysis is challenging. This paper presents an innovative methodology to handle the complexity of events in medical event logs. Based on autoencoding, accurate labels are created by clustering similar events in latent space. Moreover, the explanation of created labels is provided by the decoding of its corresponding events. Tested on synthetic events, the method is able to find hidden clusters on sparse binary data, as well as accurately explain created labels. A case study on real healthcare data is performed. Results confirm the suitability of the method to extract knowledge from complex event logs representing patient pathways.

"Bow-tie" optimal pathway discovery analysis of sepsis hospital admissions using the Hospital Episode Statistics database in England.

OBJECTIVE: The "Bow-tie" optimal pathway discovery analysis uses large clinical event datasets to map clinical pathways and to visualize risks (improvement opportunities) before, and outcomes after, a specific clinical event. This proof-of-concept study assesses the use of NHS Hospital Episode Statistics (HES) in England as a potential clinical event dataset for this pathway discovery analysis approach. MATERIALS AND METHODS: A metaheuristic optimization algorithm was used to perform the "bow-tie" analysis on HES event log data for sepsis (ICD-10 A40/A41) in 2016. Analysis of hospital episodes across inpatient and outpatient departments was performed for the period 730 days before and 365 days after the index sepsis hospitalization event. RESULTS: HES data captured a sepsis event for 76 523 individuals (>13 years), relating to 580 000 coded events (across 220 sepsis and non-sepsis event classes). The "bow-tie" analysis identified several diagnoses that most frequently preceded hospitalization for sepsis, in line with the expectation that sepsis most frequently occurs in vulnerable populations. A diagnosis of pneumonia (5 290 patients) and urinary tract infections (UTIs; 2 057 patients) most often preceded the sepsis event, with recurrent UTIs acting as a potential indicative risk factor for sepsis. DISCUSSION: This proof-of-concept study demonstrates that a "bow-tie" pathway discovery analysis of the HES database can be undertaken and provides clinical insights that, with further study, could help improve the identification and management of sepsis. The algorithm can now be more widely applied to HES data to undertake targeted clinical pathway analysis across multiple healthcare conditions.

Cost-of-illness of inflammatory bowel disease patients treated with anti-tumour necrosis factor: A French large single-centre experience.

Background: No study has evaluated the direct annual costs of inflammatory bowel disease patients treated with anti-tumour necrosis factor therapy. Objectives: The purpose of this study was to identify annual direct costs and main cost drivers of anti-tumour necrosis factor-treated inflammatory bowel disease patients. Methods: All inflammatory bowel disease patients treated with infliximab or adalimumab at Nancy University Hospital were consecutively screened for inclusion from November 2016-February 2017. Data about hospitalisation, surgery, medication, outpatient visits, investigations and transport over the previous 12 months were retrospectively collected. Results: A total of 108 patients (n = 83 Crohn's disease; n = 25 ulcerative colitis) were included. The mean annual cost per patient was €15,775 (standard deviation €7221), with no difference between Crohn's disease and ulcerative colitis (p = 0.2). The main cost driver was medication, which accounted for 84% of the total direct cost. Hospitalisation and surgery represented 11% and 2% of the direct costs. History of switch to another anti-tumour necrosis factor treatment was identified as the only independent predictor of greater direct costs in the multivariate analysis (p = 0.0018). Conclusions: In a French tertiary referral centre, direct costs of anti-tumour necrosis factor therapy-treated patients were mainly driven by medication, while hospitalisation and surgery represented only a minor part of the costs. There was no difference between Crohn's disease and ulcerative colitis patients.

Effects of Moderate Hyperbilirubinemia on Nutritive Swallowing and Swallowing-Breathing Coordination in Preterm Lambs.

BACKGROUND: Hyperbilirubinemia (HB) occurs in 90% of preterm newborns. HB induces acute neurological disorders (somnolence, abnormal tone, feeding difficulties, auditory dysfunction) and alterations in respiratory control. These findings suggest brainstem neurotoxicity that could also affect swallowing centers. OBJECTIVE: To test the hypothesis that HB impairs nutritive swallowing (NS) and swallowing-breathing coordination. METHODS: Two groups of preterm lambs (born 14 days prior to term), namely control (n = 6) and HB (n = 5), were studied. On day 5 of life (D0), moderate HB (150-250 µmol/l) was induced during 17 h in the HB group. Swallowing was assessed via recording of pharyngeal pressure and respiration by respiratory inductance plethysmography and pulse oximetry. The effect of HB on NS was assessed during standardized bottle-feeding. A second recording was performed 48 h after recovery from HB (D3). RESULTS: Swallows were less frequent (p = 0.003) and of smaller volume (p = 0.01) in HB lambs while swallowing frequency was decreased (p = 0.004). These differences disappeared after HB normalization. Swallowing-breathing coordination was impaired in HB lambs, with a decrease in percent time with NS burst-related apneas/hypopneas at D0 and D3. Simultaneously, HB lambs tended to experience more severe desaturations (<80%) during bottle-feeding. Finally, following bottle-feeding, the respiratory rate was significantly lower, along with an increased apnea duration in HB lambs. CONCLUSIONS: Swallowing and swallowing-breathing coordination are altered by acute moderate HB in preterm lambs. Decreased efficiency at bottle-feeding is accompanied by continuation of breathing during swallow bursts, which may promote lung aspiration.