The prediction of readmission or death after a hospital discharge for heart failure (HF) remains a major challenge. Modern healthcare systems, electronic health records, and machine learning (ML) ...techniques allow us to mine data to select the most significant variables (allowing for reduction in the number of variables) without compromising the performance of models used for prediction of readmission and death. Moreover, ML methods based on transformation of variables may potentially further improve the performance.
To use ML techniques to determine the most relevant and also transform variables for the prediction of 30-day readmission or death in HF patients.
We identified all Western Australian patients aged 65 years and above admitted for HF between 2003-2008 in linked administrative data. We evaluated variables associated with HF readmission or death using standard statistical and ML based selection techniques. We also tested the new variables produced by transformation of the original variables. We developed multi-layer perceptron prediction models and compared their predictive performance using metrics such as Area Under the receiver operating characteristic Curve (AUC), sensitivity and specificity.
Following hospital discharge, the proportion of 30-day readmissions or death was 23.7% in our cohort of 10,757 HF patients. The prediction model developed by us using a smaller set of variables (n = 8) had comparable performance (AUC 0.62) to the traditional model (n = 47, AUC 0.62). Transformation of the original 47 variables further improved (p<0.001) the performance of the predictive model (AUC 0.66).
A small set of variables selected using ML matched the performance of the model that used the full set of 47 variables for predicting 30-day readmission or death in HF patients. Model performance can be further significantly improved by transforming the original variables using ML methods.
Hospitalisation for atherothrombotic disease (ATD) is expected to rise in coming decades. However, increasingly, associated comorbidities impose challenges in managing patients and deciding ...appropriate secondary prevention. We investigated the prevalence and pattern of multimorbidity (presence of two or more chronic conditions) in Aboriginal and non-Aboriginal Western Australian residents with ATDs.
We used population-based de-identified linked administrative health data from 1 January 2000 to 30 June 2014 to identify a cohort of patients aged 25-59 years admitted to Western Australian hospitals with a discharge diagnosis of ATD. The prevalence of common chronic diseases in these patients was estimated and the patterns of comorbidities and multimorbidities empirically explored using two different approaches: identification of the most commonly occurring pairs and triplets of comorbid diseases, and through latent class analysis (LCA). Half of the cohort had multimorbidity, although this was much higher in Aboriginal people (Aboriginal: 79.2% vs. non-Aboriginal: 39.3%). Only a quarter were without any documented comorbidities. Hypertension, diabetes, alcohol abuse disorders and acid peptic diseases were the leading comorbidities in the major comorbid combinations across both Aboriginal and non-Aboriginal cohorts. The LCA identified four and six distinct clinically meaningful classes of multimorbidity for Aboriginal and non-Aboriginal patients, respectively. Out of the six groups in non-Aboriginal patients, four were similar to the groups identified in Aboriginal patients. The largest proportion of patients (33% in Aboriginal and 66% in non-Aboriginal) was assigned to the "minimally diseased" (or relatively healthy) group, with most patients having less than two conditions. Other groups showed variability in degree and pattern of multimorbidity.
Multimorbidity is common in ATD patients and the comorbidities tend to interact and cluster together. Physicians need to consider these in their clinical practice. Different treatment and secondary prevention strategies are likely to be useful for management in these cluster groups.
This study explores the effectiveness of Explainable Artificial Intelligence (XAI) for predicting suicide risk from medical tabular data. Given the common challenge of limited datasets in ...health-related Machine Learning (ML) applications, we use data augmentation in tandem with ML to enhance the identification of individuals at high risk of suicide. We use SHapley Additive exPlanations (SHAP) for XAI and traditional correlation analysis to rank feature importance, pinpointing primary factors influencing suicide risk and preventive measures. Experimental results show the Random Forest (RF) model is excelling in accuracy, F1 score, and AUC (>97% across metrics). According to SHAP, anger issues, depression, and social isolation emerge as top predictors of suicide risk, while individuals with high incomes, esteemed professions, and higher education present the lowest risk. Our findings underscore the effectiveness of ML and XAI in suicide risk assessment, offering valuable insights for psychiatrists and facilitating informed clinical decisions.
Osteoporosis is a common condition associated with fragility fractures, especially in older individuals and women. Antidepressants have emerged as a potential risk factor, but their association with ...bone fragility remains uncertain because the results of past studies are difficult to generalize. We aimed to investigate the association between antidepressant exposure and subsequent treatment for osteoporosis in a nationally representative sample of Australians.
Cohort study using a 10% random sample of the Pharmaceutical Benefits Scheme (PBS) data for 2012, that included 566,707 individuals aged older than or equal to 50 years not dispensed osteoporosis medications. The effect of exposure to antidepressants during 2012 (prevalent or incident) or later (up to 2022) was examined using Cox regression models adjusted for age, sex, comorbidities and other psychotropic medications.
Over 10 years, 73,360 (12.94%) received osteoporosis medications; 16,216 (22.10%) had been dispensed antidepressants in 2012. The hazard of osteoporosis medication dispensing was higher among those exposed to antidepressants (HR = 1.16, 99% CI = 1.14-1.18; average duration of follow up: 8.0 ± 3.1 years, range: 1-10 years). The hazard of osteoporosis medication diminished with increasing age, and the effect of antidepressants was 37%-76% more pronounced among men in the 50s and 60s. Different classes of antidepressants had a similar risk profile.
The dispensing of antidepressants in older age is associated with higher hazard of subsequent dispensing of medications for osteoporosis, and this association is more marked for young older adults, particularly men. Clinicians should monitor the bone health of older individuals treated with antidepressants in order to decrease the morbidity associated with fragility fractures.
There are no overviews of systematic reviews investigating haemoglobin thresholds for transfusion. This is important as the literature on transfusion thresholds has grown considerably in recent ...years. Our aim was to synthesise evidence from systematic reviews and meta-analyses of the effects of restrictive and liberal transfusion strategies on mortality.
This was a systematic review of systematic reviews (overview). We searched MEDLINE, Embase, Web of Science Core Collection, PubMed, Google Scholar, and the Joanna Briggs Institute EBP Database, from 2008 to 2018. We included systematic reviews and meta-analyses of randomised controlled trials comparing mortality in patients assigned to red cell transfusion strategies based on haemoglobin thresholds. Two independent reviewers extracted data and assessed methodological quality. We assessed the methodological quality of included reviews using AMSTAR 2 and the quality of evidence pooled using an algorithm to assign GRADE levels.
We included 19 systematic reviews reporting 33 meta-analyses of mortality outcomes from 53 unique randomised controlled trials. Of the 33 meta-analyses, one was graded as high quality, 15 were moderate, and 17 were low. Of the meta-analyses presenting high- to moderate-quality evidence, 12 (75.0%) reported no statistically significant difference in mortality between restrictive and liberal transfusion groups and four (25.0%) reported significantly lower mortality for patients assigned to a restrictive transfusion strategy. We found few systematic reviews addressed clinical differences between included studies: variation was observed in haemoglobin threshold concentrations, the absolute between group difference in haemoglobin threshold concentration, time to randomisation (resulting in transfusions administered prior to randomisation), and transfusion dosing regimens.
Meta-analyses graded as high to moderate quality indicate that in most patient populations no difference in mortality exists between patients assigned to a restrictive or liberal transfusion strategy.
PROSPERO CRD42019120503.
Background In 2018, the World Health Organization prioritized control of acute rheumatic fever (ARF) and rheumatic heart disease (RHD), including disease surveillance. We developed strategies for ...estimating contemporary ARF/RHD incidence and prevalence in Australia (2015-2017) by age group, sex, and region for Indigenous and non-Indigenous Australians based on innovative, direct methods. Methods and Results This population-based study used linked administrative data from 5 Australian jurisdictions. A cohort of ARF (age <45 years) and RHD cases (<55 years) were sourced from jurisdictional ARF/RHD registers, surgical registries, and inpatient data. We developed robust methods for epidemiologic case ascertainment for ARF/RHD. We calculated age-specific and age-standardized incidence and prevalence. Age-standardized rate and prevalence ratios compared disease burden between demographic subgroups. Of 1425 ARF episodes, 72.1% were first-ever, 88.8% in Indigenous people and 78.6% were aged <25 years. The age-standardized ARF first-ever rates were 71.9 and 0.60/100 000 for Indigenous and non-Indigenous populations, respectively (age-standardized rate ratio=124.1; 95% CI, 105.2-146.3). The 2017 Global Burden of Disease RHD prevalent counts for Australia (<55 years) underestimate the burden (1518 versus 6156 Australia-wide extrapolated from our study). The Indigenous age-standardized RHD prevalence (666.3/100 000) was 61.4 times higher (95% CI, 59.3-63.5) than non-Indigenous (10.9/100 000). Female RHD prevalence was double that in males. Regions in northern Australia had the highest rates. Conclusions This study provides the most accurate estimates to date of Australian ARF and RHD rates. The high Indigenous burden necessitates urgent government action. Findings suggest RHD may be underestimated in many high-resource settings. The linked data methods outlined here have potential for global applicability.
•A comparison of multiple tree-based machine learning classifiers for predicting acute coronary syndrome from administrative data.•A quantified analysis of multiple explainable artificial ...intelligence approaches.•A ranking of the patient features (including drug and comorbidity history) that are most important when predicting acute coronary syndrome.•An explainable artificial intelligence based method for predicting drug safety for pharmacovigilance.
Background and Objective. Explainable Artificial Intelligence (XAI) has been identified as a viable method for determining the importance of features when making predictions using Machine Learning (ML) models. In this study, we created models that take an individual’s health information (e.g. their drug history and comorbidities) as inputs, and predict the probability that the individual will have an Acute Coronary Syndrome (ACS) adverse outcome. Methods. Using XAI, we quantified the contribution that specific drugs had on these ACS predictions, thus creating an XAI-based technique for pharmacovigilance monitoring, using ACS as an example of the adverse outcome to detect. Individuals aged over 65 who were supplied Musculo-skeletal system (anatomical therapeutic chemical (ATC) class M) or Cardiovascular system (ATC class C) drugs between 1993 and 2009 were identified, and their drug histories, comorbidities, and other key features were extracted from linked Western Australian datasets. Multiple ML models were trained to predict if these individuals would have an ACS related adverse outcome (i.e., death or hospitalisation with a discharge diagnosis of ACS), and a variety of ML and XAI techniques were used to calculate which features — specifically which drugs — led to these predictions. Results. The drug dispensing features for rofecoxib and celecoxib were found to have a greater than zero contribution to ACS related adverse outcome predictions (on average), and it was found that ACS related adverse outcomes can be predicted with 72% accuracy. Furthermore, the XAI libraries LIME and SHAP were found to successfully identify both important and unimportant features, with SHAP slightly outperforming LIME. Conclusions. ML models trained on linked administrative health datasets in tandem with XAI algorithms can successfully quantify feature importance, and with further development, could potentially be used as pharmacovigilance monitoring techniques.
Real-world evidence is limited on whether antihypertensive medications help avert major adverse cardiovascular events (MACE) after stroke without increasing the risk of falls. We investigated the ...association of adherence to antihypertensive medications on the incidence of MACE and falls requiring hospitalization after stroke.
A retrospective cohort study of adults who were newly dispensed antihypertensive medications after an acute stroke (Australian Stroke Clinical Registry 2012-2016; Queensland and Victoria). Pharmaceutical dispensing records were used to determine medication adherence according to the proportion of days covered in the first 6 months poststroke. Outcomes between 6 and 18 months postdischarge included: (i) MACE, a composite outcome of all-cause death, recurrent stroke or acute coronary syndrome; and (ii) falls requiring hospitalization. Estimates were derived using Cox models, adjusted for >30 confounders using inverse probability treatment weights.
Among 4076 eligible participants (median age 68 years; 37% women), 55% had a proportion of days covered ≥80% within 6 months postdischarge. In the subsequent 12 months, 360 (9%) participants experienced a MACE and 337 (8%) experienced a fall requiring hospitalization. After achieving balance between groups, participants with a proportion of days covered ≥80% had a reduced risk of MACE (hazard ratio: 0.68; 95% CI: 0.54-0.84) and falls requiring hospitalization (subdistribution hazard ratio: 0.78; 95% CI: 0.62-0.98) than those with a proportion of days covered <80%.
High adherence to antihypertensive medications within 6 months poststroke was associated with reduced risks of both MACE and falls requiring hospitalization. Patients should be encouraged to adhere to their antihypertensive medications to maximize poststroke outcomes.
Health care disparity is a public health challenge. We compared the prevalence of diabetes, quality of care and outcomes between mental health clients (MHCs) and non-MHCs.
This was a population-based ...longitudinal study of 139,208 MHCs and 294,180 matched non-MHCs in Western Australia (WA) from 1990 to 2006, using linked data of mental health registry, electoral roll registrations, hospital admissions, emergency department attendances, deaths, and Medicare and pharmaceutical benefits claims. Diabetes was identified from hospital diagnoses, prescriptions and diabetes-specific primary care claims (17,045 MHCs, 26,626 non-MHCs). Both univariate and multivariate analyses adjusted for socio-demographic factors and case mix were performed to compare the outcome measures among MHCs, category of mental disorders and non-MHCs.
The prevalence of diabetes was significantly higher in MHCs than in non-MHCs (crude age-sex-standardised point-prevalence of diabetes on 30 June 2006 in those aged ≥20 years, 9.3% vs 6.1%, respectively, P < 0.001; adjusted odds ratio (OR) 1.40, 95% CI 1.36 to 1.43). Receipt of recommended pathology tests (HbA1c, microalbuminuria, blood lipids) was suboptimal in both groups, but was lower in MHCs (for all tests combined; adjusted OR 0.81, 95% CI 0.78 to 0.85, at one year; and adjusted rate ratio (RR) 0.86, 95% CI 0.84 to 0.88, during the study period). MHCs also had increased risks of hospitalisation for diabetes complications (adjusted RR 1.20, 95% CI 1.17 to 1.24), diabetes-related mortality (1.43, 1.35 to 1.52) and all-cause mortality (1.47, 1.42 to 1.53). The disparities were most marked for alcohol/drug disorders, schizophrenia, affective disorders, other psychoses and personality disorders.
MHCs warrant special attention for primary and secondary prevention of diabetes, especially at the primary care level.
Effectiveness of E-learning in Pharmacy Education Salter, Sandra M.; Karia, Ajay; Sanfilippo, Frank M. ...
American journal of pharmaceutical education,
05/2014, Volume:
78, Issue:
4
Journal Article
Peer reviewed
Open access
Over the past 2 decades, e-learning has evolved as a new pedagogy within pharmacy education. As learners and teachers increasingly seek e-learning opportunities for an array of educational and ...individual benefits, it is important to evaluate the effectiveness of these programs. This systematic review of the literature examines the quality of e-learning effectiveness studies in pharmacy, describes effectiveness measures, and synthesizes the evidence for each measure. E-learning in pharmacy education effectively increases knowledge and is a highly acceptable instructional format for pharmacists and pharmacy students. However, there is limited evidence that e-learning effectively improves skills or professional practice. There is also no evidence that e-learning is effective at increasing knowledge long term; thus, long-term follow-up studies are required. Translational research is also needed to evaluate the benefits of e-learning at patient and organizational levels.