Machine learning algorithms to predict colistin-induced nephrotoxicity from electronic health records in patients with multidrug-resistant Gram-negative infection.

OBJECTIVES: Colistin-induced nephrotoxicity prolongs hospitalisation and increases mortality. The study aimed to construct machine learning models to predict colistin-induced nephrotoxicity in patients with multidrug-resistant Gram-negative infection. METHODS: Patients receiving colistin from three hospitals in the Clinical Research Database were included. Data were divided into a derivation cohort (2011-2017) and a temporal validation cohort (2018-2020). Fifteen machine learning models were established by categorical boosting, light gradient boosting machine and random forest. Classifier performances were compared by the sensitivity, F1 score, Matthews correlation coefficient (MCC), area under the receiver operating characteristic (AUROC) curve, and area under the precision-recall curve (AUPRC). SHapley Additive exPlanations plots were drawn to understand feature importance and interactions. RESULTS: The study included 1392 patients, with 360 (36.4%) and 165 (40.9%) experiencing nephrotoxicity in the derivation and temporal validation cohorts, respectively. The categorical boosting with oversampling achieved the highest performance with a sensitivity of 0.860, an F1 score of 0.740, an MCC of 0.533, an AUROC curve of 0.823, and an AUPRC of 0.737. The feature importance demonstrated that the days of colistin use, cumulative dose, daily dose, latest C-reactive protein, and baseline haemoglobin were the most important risk factors, especially for vulnerable patients. A cutoff colistin dose of 4.0 mg/kg body weight/d was identified for patients at higher risk of nephrotoxicity. CONCLUSIONS: Machine learning techniques can be an early identification tool to predict colistin-induced nephrotoxicity. The observed interactions suggest a modification in dose adjustment guidelines. Future geographic and prospective validation studies are warranted to strengthen the real-world applicability.

Efficacy of Osteoporosis Medications for Patients With Chronic Kidney Disease: An Updated Systematic Review and Network Meta-Analysis.

Background: Chronic kidney disease (CKD) is associated with bone and mineral metabolism. In this study we evaluated the comparative efficacies and safety of osteoporosis medications in patients with CKD or a history of kidney transplantation, and make recommendations for the best choice of osteoporosis treatment among patients with CKD or a history of kidney transplantation. Methods: We systemically searched for randomized controlled trials published in PubMed, Embase, and Cochrane databases up to June 2020. Network-meta analysis was used to compare the relative effectiveness of different treatments. A random-effects model was used when heterogeneity was expected. The safety of different treatments was also evaluated in terms of reported major adverse events. Results: A total of 17 studies with data from 10,214 patients who had stage 2-5 CKD, were receiving dialysis, or had a history of kidney transplantation were included in the network meta-analysis. Treatment with teriparatide, denosumab, alendronate, and raloxifene were all associated with a significantly reduced risk of fractures compared to treatment with placebos [teriparatide: odds ratio (OR) = 0.19, 95% confidence interval (CI): 0.10-0.35; denosumab: OR = 0.40, 95% CI: 0.27-0.58; alendronate: OR = 0.61, 95% CI: 0.40-0.92; raloxifene: OR = 0.52, 95% CI: 0.41-0.67]. The rank probability and the surface under the cumulative ranking (SUCRA) values suggested that teriparatide ranked the highest for improvement in vertebral bone mineral density (BMD) (SUCRA = 97.8%), whereas denosumab ranked the highest for improvement in femoral neck BMD (SUCRA = 88.3%). Conclusion: Teriparatide and denosumab seem to be the most effective treatments for preventing bone loss and reducing the risk of fracture in our network comparison. However, because of the limitations and potential biases in the reviewed studies, there is still some uncertainty about the best treatment options for osteoporosis in patients with CKD or a history of kidney transplantation. Systematic Review Registration: [PROSPERO], identifier [CRD42020209830].

A qualitative study on the development of pharmacist-managed clinics in Taiwan.

WHAT IS KNOWN AND OBJECTIVE: Pharmacist-managed clinics (PMCs) are established to solve drug-related problems and enhance the quality of care of ambulatory patients. Although the benefits of such services have been demonstrated, little is known about PMC operations, especially outside the United States. The aim of this study was to explore how PMCs were established and to discuss implementation issues of PMCs in Taiwan. METHODS: A purposive sample of pharmacists, pharmacy administrators and physicians involved with PMCs was recruited from hospitals of varying scales across Taiwan. Semi-structured, individual interviews were conducted to understand the perceptions of the clinical service of PMCs. Interviews were transcribed verbatim and analysed by thematic analysis to find underlying themes. RESULTS: A total of 12 pharmacists, 5 pharmacy administrators and 3 physicians from 8 institutions were interviewed. Pharmacists spent 4 to 20 h per week at PMCs, and the practice experiences of PMC ranged from 1 to 6 years. PMCs have been provided in these institutions for 4 to 11 years with an average volume of 28 h and 25 patient visits weekly. Study participants described influential factors in establishing PMCs, including clinical expertise, attitude towards patient care and trust building with collaborating physicians. Operational concerns in implementing PMCs included role clarifications, manpower shortage, inadequate advanced training or certification, regulatory issues and a lack of service promotion. WHAT IS NEW AND CONCLUSION: This research broadens the understanding of operating PMC services and reveals key requirements and concerns regarding the care model, which can be useful for other countries. Resolving perceived barriers and collecting other stakeholders' perspectives may reinforce the integration of PMCs into patient care in the future.

Outflow tract geometries are associated with adverse outcome indicators in repaired tetralogy of Fallot.

OBJECTIVES: A wide variety of right ventricular outflow tract (RVOT) and pulmonary artery (PA) geometries has been reported in patients with repaired tetralogy of Fallot (rTOF). We aimed to investigate the associations between RVOT/PA geometries and outcome indicators in a large rTOF cohort receiving non-conduit repair. METHODS: Three-dimensional magnetic resonance angiographic images of 206 patients with rTOF who had a pulmonary regurgitation (PR) fraction ≥20% were reviewed. Patients' RVOT geometry was quantitatively classified into 4 distinct shapes (tubular, hourglass, pyramid, and inverted trapezoid). Bilateral PA size discrepancy was defined as the diameter of the smaller side being less than 70% of that of the bigger side. RESULTS: Based on lateral projection of the 3-dimensional images, patients with an inverted trapezoid-shaped RVOT had the smallest RV end-diastolic volume index (EDVi) (108.7 ± 24.3 mL/m2) and pulmonary valve annulus diameter, and shortest QRS duration, whereas those with a pyramid-shaped RVOT had the largest RV EDVi (161.0 ± 44.6 mL/m2) and pulmonary valve annulus diameter. Similar trends of differences were also observed if such classifications were based on the frontal projections. Multivariable analysis revealed that RVOT shapes, subvalvular diameter, PR fraction, QRS duration, and the presence of bilateral PA size discrepancy were independent determinants of RV EDVi. Furthermore, having bilateral PA size discrepancy (25.2%) was independently associated with lower peak oxygen consumption (P = .041). CONCLUSIONS: Distinct RVOT morphologies and branch PA size discrepancy are associated with variations in RV remodeling and exercise capacity in patients with rTOF. These findings may aid decision-making regarding reintervention for PR and branch PA size discrepancy.

Comparison of the predictive outcomes for anti-tuberculosis drug-induced hepatotoxicity by different machine learning techniques.

BACKGROUND: The study compared the predictive outcomes of artificial neural network, support vector machine and random forest on the occurrence of anti-tuberculosis drug-induced hepatotoxicity. METHODS: The clinical and genomic data of patients treated with anti-tuberculosis drugs at Taipei Medical University-Wanfang Hospital were used as training sets, and those at Taipei Medical University-Shuang Ho Hospital served as test sets. Features were selected through a univariate risk factor analysis and literature evaluation. The accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve were calculated to compare the traditional, genomic, and combined models of the three techniques. RESULTS: Nine models were created with 7 clinical factors and 4 genotypes. Artificial neural network with clinical and genomic factors exhibited the best performance, with an accuracy of 88.67%, a sensitivity of 80%, and a specificity of 90.4% for the test set. The area under the receiver operating characteristic curve of this best model reached 0.894 for training set and 0.898 for test set, which was significantly better than 0.801 for training set and 0.728 for test set by support vector machine and 0.724 for training set and 0.718 for test set by random forest. CONCLUSIONS: Artificial neural network with clinical and genomic data can become a clinical useful tool in predicting anti-tuberculosis drug-induced hepatotoxicity. The machine learning technique can be an innovation to predict and prevent adverse drug reaction.

Effects and Safety of an Oral Adsorbent on Chronic Kidney Disease Progression: A Systematic Review and Meta-Analysis.

BACKGROUND: AST-120 (Kremezin), which is an oral spherical carbon adsorbent, has been reported to have the potential for retarding disease progression in patients with chronic kidney disease. We aimed to evaluate its efficacy and safety in this study. METHODS: We systematically searched for randomized controlled trials published in PubMed, Embase, and Cochrane databases. The primary outcomes were the renal outcome and all-cause mortality, and the change in serum indoxyl sulfate (IS) levels. The safety outcome was also evaluated in terms of reported major adverse events. A random-effects model was used when heterogeneity was expected. RESULTS: Eight studies providing data for 3349 patients were included in the meta-analysis. The risk ratio of renal outcome and all-cause mortality were 0.97 (95% CI: 0.88-1.07; 6 trials) and 0.94 (0.73-1.20; 5 trials), respectively. Furthermore, the weighted mean change in IS levels from baseline to the end of the study was -0.28 mg/dL (95% CI: -0.46 to -0.11; 4 trials). CONCLUSIONS: This study provides evidence that AST-120 can effectively lower IS levels but still controversial in terms of slowing disease progression and all-cause mortality. Except for dermatological events, the incidence of adverse events did not differ significantly between the AST-120 and placebo groups.

Treatment with normobaric or hyperbaric oxygen and its effect on neuropsychometric dysfunction after carbon monoxide poisoning: A systematic review and meta-analysis of randomized controlled trials.

BACKGROUND: Carbon monoxide (CO) poisoning may result in acute neurological sequelae, cognitive sequelae, and delay neurological sequelae. The administration of hyperbaric oxygen (HBO) to prevent the development of delayed neurological sequelae in CO poisoning have extensively investigated but conflicting results have been reported. We performed a systematic literature review and meta-analysis of randomized controlled trials (RCTs) evaluating HBO treatment and its effect on neuropsychometric dysfunction after CO poisoning. METHODS: We searched Medline, Embase, Pubmed, and the Cochrane Register of Controlled Trials from inception to December 2017. Eligible studies compared HBO therapy with normobaric oxygen (NBO) in patients with CO poisoning. RESULTS: Six studies compared HBO with NBO in CO poisoning patients. Compared with patients treated with NBO, a lower percentage of patients treated with HBO reported headache (16.2% vs 16.5%, relative risk [RR] = 0.83, 95% CI = 0.38-1.80), memory impairment (18.2% vs 23.8%, RR = 0.80, 95% CI = 0.43-1.49), difficulty concentrating (15.0% vs 18.4%, RR = 0.86, 95% CI = 0.55-1.34), and disturbed sleep (14.7% vs 16.2%, RR = 0.91, 95% CI = 0.59-1.39). Two sessions of HBO treatment exhibited no advantage over one session. CONCLUSIONS: The meta-analysis indicated that compared with CO poisoning patients treated with NBO, HBO treated patients have a lower incidence of neuropsychological sequelae, including headache, memory impairment, difficulty concentrating, disturbed sleep, and delayed neurological sequelae. Taking into consideration the cost-effectiveness of one session of HBO, one session of HBO treatment could be an economical option for patients with CO poisoning with high severity.

Efficacy and Safety of Rituximab for the Treatment of Graves' Orbitopathy: A Meta-analysis of Randomized Controlled Trials.

STUDY OBJECTIVE: To investigate the efficacy and safety of rituximab in patients with Graves' orbitopathy (GO). DESIGN: Systematic review and meta-analysis of four randomized controlled trials. PATIENTS: A total of 293 patients with GO who received rituximab or control (either glucocorticoids, the established first-line therapy [three trials], or saline [one trial]). MEASUREMENTS AND RESULTS: Relevant studies published before February 2018 were identified from the PubMed, EMBASE, Cochrane Library, and Scopus databases and the ClinicalTrials.gov registry. Individual effect sizes were standardized, and a meta-analysis was conducted to calculate the pooled effect size by using a random-effects model. Treatment efficacy was assessed by measuring the following outcomes: clinical activity score (CAS), sight visual acuity reduction (NOSPECS) score, proptosis, diplopia, changes in eye volume, quality of life, and adverse events. In the four included trials, 113 patients in the rituximab group and 108 patients in the control group were evaluated. Compared with the control group, CAS (weighted mean difference 0.57, 95% confidence interval 0.25-0.89) was significantly reduced at 24 weeks in the rituximab group. Compared with the control group, considerable proptosis reduction was also observed in the rituximab group; however, the difference was not significant. The proportion of adverse events in the rituximab group was not significantly higher than that in the glucocorticoid control group, but one of the included trials indicated that the rituximab group had more serious adverse events than the saline control group. CONCLUSION: Rituximab is a relatively safe and viable treatment that is superior to glucocorticoids or saline for patients with moderate to severe GO. However, the incidence of serious adverse events was disparate among the included trials. Additional studies involving a larger sample size and investigating the optimal rituximab dosage, frequency, and method of administration are warranted.

Preliminary physician and pharmacist survey of the National Health Insurance PharmaCloud system in Taiwan.

BACKGROUND: The PharmaCloud system, a cloud-based medication system, was launched by the Taiwan National Health Insurance Administration (NHIA) in 2013 to integrate patients' medication lists among different medical institutions. The aim of the preliminary study was to evaluate satisfaction with this system among physicians and pharmacists at the early stage of system implementation. METHODS: A questionnaire was developed through a review of the literature and discussion in 6 focus groups to understand the level of satisfaction, attitudes, and intentions of physicians and pharmacists using the PharmaCloud system. It was then administered nationally in Taiwan in July to September 2015. Descriptive statistics and multiple regression were performed to identify variables influencing satisfaction and intention to use the system. RESULTS: In total, 895 pharmacist and 105 physician questionnaires were valid for analysis. The results showed that satisfaction with system quality warranted improvement. Positive attitudes toward medication reconciliation among physicians and pharmacists, which were significant predictors of the intention to use the system (ß= 0.223, p < 0.001). Most physicians and pharmacists agreed that obtaining signed patient consent was needed but preferred that it be conducted by the NHIA rather than by individual medical institutions (4.02 ± 1.19 vs. 3.49 ± 1.40, p < 0.01). CONCLUSIONS: The preliminary study results indicated a moderate satisfaction toward the PharmaCloud system. Hospital pharmacists had a high satisfaction rate, but neither are physicians and community pharmacists. Continuously improvement on system quality has been performing based on the results of this preliminary survey. Policies and standardization processes, including privacy protection, are still warranted further actions to make the Taiwan PharmaCloud system a convenient platform for medication reconciliation.

The effect of medication therapy management service combined with a national PharmaCloud system for polypharmacy patients.

BACKGROUND: This study evaluated a medication therapy management service using the Taiwan National Health Insurance Administration's PharmaCloud system in a medical center in Taiwan. The new PharmaCloud System, launched in 2013, links a complete list of prescribed and dispensed medication from different hospitals, clinics, and pharmacies for all insured patients. METHOD: The study included patients with polypharmacy (≥5 drugs) at a medication therapy management service from March 2013 to March 2014. A structured questionnaire was designed to collect patients' baseline data and record patients' knowledge, attitudes, and practice scores before and after the service intervention. Phone follow-ups for practice and adherence scores on medication use were performed after 3 months. RESULTS: There were 152 patients recruited in the study. Scores for medication use attitudes and practice significantly increased after the service (attitudes: 40.06 ± 0.26 to 43.07 ± 0.19, p <0.001; practice: 33.42 ± 0.30 to 40.37 ± 0.30, p <0.001). The scores for medication adherence also increased from 3.02 ± 0.07 to 3.92 ± 0.02 (p <0.001). CONCLUSIONS: The PharmaCloud system facilitates accurate and efficient medication reconciliation for pharmacists in the medication therapy management service. The model improved patients' attitudes and practice of the rational use of medications and adherence with medications. Further studies are warranted to evaluate human resources, executing costs, and the cost-benefit ratio of this medication therapy management service with the PharmaCloud system.

Nephrotoxicity of vancomycin in patients with methicillin-resistant Staphylococcus aureus bacteraemia.

AIM: Vancomycin and teicoplanin are the two most used glycopeptides for the treatment of methicillin-resistant Staphylococcus aureus (MRSA). Vancomycin is suspected to have more nephrotoxicity but this has not been clearly established. The aim of this study was to assess its nephrotoxicity by a consensus definition of acute kidney injury (AKI): the risk (R), injury (I), failure (F), loss and end-stage renal disease (RIFLE) classification. METHODS: Patients with MRSA bacteraemia who were prescribed either vancomycin or teicoplanin between 2003 and 2008 were classified. Patients who developed AKI were classified by RIFLE criteria. Variables such as comorbidities, laboratory data and medical cost information were also obtained from the database. Outcomes determined were: (i) the rate of nephrotoxicity and mortality; and (ii) the association of nephrotoxicity with the length of hospital stay and costs. RESULTS: The study included 190 patients (vancomycin 33, teicoplanin 157). Fifteen patients on vancomycin and 27 patients on teicoplanin developed AKI (P = 0.0004). In the vancomycin group, four, eight and three patients were classified to RIFLE criteria R, I and F, respectively. In the teicoplanin group, 17, nine and one patient were classified to RIFLE criteria R, I and F, respectively. Kaplan-Meier analysis showed significant difference in time to nephrotoxicity for the vancomycin group compared to the teicoplanin group. No significant differences were found between the groups in terms of total mortality, length of hospital stay and costs. CONCLUSION: The study data suggest that vancomycin is associated with a higher likelihood of nephrotoxicity using the RIFLE classification.