Using machine learning to examine medication adherence thresholds and risk of hospitalization

Exploring patient medication adherence and data mining methods in clinical big data: A contemporary review

BACKGROUND

Increasingly, patient medication adherence data are being consolidated from claims databases and electronic health records (EHRs). Such databases offer an indirect avenue to gauge medication adherence in our data-rich healthcare milieu. The surge in data accessibility, coupled with the pressing need for its conversion to actionable insights, has spotlighted data mining, with machine learning (ML) emerging as a pivotal technique. Nonadherence poses heightened health risks and escalates medical costs. This paper elucidates the synergistic interaction between medical database mining for medication adherence and the role of ML in fostering knowledge discovery.

METHODS

We conducted a comprehensive review of EHR applications in the realm of medication adherence, leveraging ML techniques. We expounded on the evolution and structure of medical databases pertinent to medication adherence and harnessed both supervised and unsupervised ML paradigms to delve into adherence and its ramifications.

RESULTS

Our study underscores the applications of medical databases and ML, encompassing both supervised and unsupervised learning, for medication adherence in clinical big data. Databases like SEER and NHANES, often underutilized due to their intricacies, have gained prominence. Employing ML to excavate patient medication logs from these databases facilitates adherence analysis. Such findings are pivotal for clinical decision-making, risk stratification, and scholarly pursuits, aiming to elevate healthcare quality.

CONCLUSION

Advanced data mining in the era of big data has revolutionized medication adherence research, thereby enhancing patient care. Emphasizing bespoke interventions and research could herald transformative shifts in therapeutic modalities.

Identification of high-risk patients for referral through machine learning assisting the decision making to manage minor ailments in community pharmacies

Background: Data analysis techniques such as machine learning have been used for assisting in triage and the diagnosis of health problems. Nevertheless, it has not been used yet to assist community pharmacists with services such as the Minor Ailment Services These services have been implemented to reduce the burden of primary care consultations in general medical practitioners (GPs) and to allow a better utilization of community pharmacists' skills. However, there is a need to refer high-risk patients to GPs. Aim: To develop a predictive model for high-risk patients that need referral assisting community pharmacists' triage through a minor ailment service. Method: An ongoing pragmatic type 3 effectiveness-implementation hybrid study was undertaken at a national level in Spanish community pharmacies since October 2020. Pharmacists recruited patients presenting with minor ailments and followed them 10 days after the consultation. The main outcome measured was appropriate medical referral (in accordance with previously co-designed protocols). Nine machine learning models were tested (three statistical, three black box and three tree models) to assist pharmacists in the detection of high-risk individuals in need of referral. Results: Over 14'000 patients were included in the study. Most patients were female (68.1%). With no previous treatment for the specific minor ailment (68.0%) presented. A percentage of patients had referral criteria (13.8%) however, not all of these patients were referred by the pharmacist to the GP (8.5%). The pharmacists were using their clinical expertise not to refer these patients. The primary prediction model was the radial support vector machine (RSVM) with an accuracy of 0.934 (CI95 = [0.926,0.942]), Cohen's kappa of 0.630, recall equal to 0.975 and an area under the curve of 0.897. Twenty variables (out of 61 evaluated) were included in the model. radial support vector machine could predict 95.2% of the true negatives and 74.8% of the true positives. When evaluating the performance for the 25 patient's profiles most frequent in the study, the model was considered appropriate for 56% of them. Conclusion: A RSVM model was obtained to assist in the differentiation of patients that can be managed in community pharmacy from those who are at risk and should be evaluated by GPs. This tool potentially increases patients' safety by increasing pharmacists' ability to differentiate minor ailments from other medical conditions.

Determinants of adherence to treatment in patients with ophthalmic conditions

BACKGROUND

The objective of this study was to identify and determine factors associated with patients' ophthalmic adherence in common ocular conditions from randomized clinical trials (RCT).

RESEARCH DESIGN METHODS

A univariate analysis with proportions, a bivariate analysis using polychoric correlations, and logistic regression (LR) models were used. The collected dataset was made up of records from RCT. Using three validated LR models, factors were identified and ranked based on their adjusted odds ratio and their statistical significance to adherence.

RESULTS

A total of 1,087 valid patients were included in this analysis, of which 88.96% presented adherence. All models were calibrated, had a good performance, were well specified and cost-effective using the Hosmer-Lemeshow test, metrics for class imbalance, link test approach and Akaike's criteriums, respectively.

CONCLUSION

We identified as determinants for encouraging good ophthalmic adherence the adverse events presented, duration of the study, female sex, and older age; other determinants such as medical condition, protocol treatment, type of treatment and disease are all risk factors for adherence. Improvements in ophthalmic adherence may be achieved by focused attention to young male patients with chronic degenerative diseases such as glaucoma or ocular hypertension (especially those who need combination therapy) and developing medications with reduced side effects.

Adherence to Oral Antidiabetic Drugs in Patients with Type 2 Diabetes: Systematic Review and Meta-Analysis

Poor adherence to oral antidiabetic drugs (OADs) in patients with type 2 diabetes (T2D) can lead to therapy failure and risk of complications. The aim of this study was to produce an adherence proportion to OADs and estimate the association between good adherence and good glycemic control in patients with T2D. We searched in MEDLINE, Scopus, and CENTRAL databases to find observational studies on therapeutic adherence in OAD users. We calculated the proportion of adherent patients to the total number of participants for each study and pooled study-specific adherence proportions using random effect models with Freeman-Tukey transformation. We also calculated the odds ratio (OR) of having good glycemic control and good adherence and pooled study-specific OR with the generic inverse variance method. A total of 156 studies (10,041,928 patients) were included in the systematic review and meta-analysis. The pooled proportion of adherent patients was 54% (95% confidence interval, CI: 51-58%). We observed a significant association between good glycemic control and good adherence (OR: 1.33; 95% CI: 1.17-1.51). This study demonstrated that adherence to OADs in patients with T2D is sub-optimal. Improving therapeutic adherence through health-promoting programs and prescription of personalized therapies could be an effective strategy to reduce the risk of complications.

Medication Adherence and Cardiometabolic Control Indicators Among American Indian Adults Receiving Tribal Health Services: Protocol for a Longitudinal Electronic Health Records Study

BACKGROUND

American Indian adults have the highest prevalence of type 2 diabetes (T2D) in any racial or ethnic group and experience high rates of comorbidities. Uncontrolled cardiometabolic risk factors-insulin resistance, resulting in impaired glucose tolerance, dyslipidemia, and hypertension-increase the risk of mortality. Mortality is significantly reduced by glucose- and lipid-lowering and antihypertensive medication adherence. Medication adherence is low among American Indian adults living in non-Indian Health Service health care settings. Virtually nothing is known about the nature and extent of medication adherence among reservation-dwelling American Indian adults who primarily receive their medications without cost from Indian Health Service or tribal facilities. Electronic health records (EHRs) offer a rich but underused data source regarding medication adherence and its potential to predict cardiometabolic control indicators (C-MCIs). With the support of the Choctaw Nation of Oklahoma (CNO), we address this oversight by using EHR data generated by this large, state-of-the-art tribal health care system to investigate C-MCIs.

OBJECTIVE

Our specific aims are to determine, using 2018 EHR data, the bivariate relationships between medication adherence and C-MCIs, demographics, and comorbidities and each C-MCI and demographics and comorbidities; develop machine learning models for predicting future C-MCIs from the previous year's medication adherence, demographics, comorbidities, and common laboratory tests; and identify facilitators of and barriers to medication adherence within the context of social determinants of health (SDOH), EHR-derived medication adherence, and C-MCIs.

METHODS

Drawing on the tribe's EHR (2018-2021) data for CNO patients with T2D, we will characterize the relationships among medication adherence (to glucose- and lipid-lowering and antihypertensive drugs) and C-MCIs (hemoglobin A1c ≤7%, low-density lipoprotein cholesterol <100 mg/dL, and systolic blood pressure <130 mm Hg); patient demographics (eg, age, sex, SDOH, and residence location); and comorbidities (eg, BMI ≥30, cardiovascular disease, and chronic kidney disease). We will also characterize the association of each C-MCI with demographics and comorbidities. Prescription and pharmacy refill data will be used to calculate the proportion of days covered with medications, a typical measure of medication adherence. Using machine learning techniques, we will develop prediction models for future (2019-2021) C-MCIs based on medication adherence, patient demographics, comorbidities, and common laboratory tests (eg, lipid panel) from the previous year. Finally, key informant interviews (N=90) will explore facilitators of and barriers to medication adherence within the context of local SDOH.

RESULTS

Funding was obtained in early 2022. The University of Florida and CNO approved the institutional review board protocols and executed the data use agreements. Data extraction is in process. We expect to obtain results from aims 1 and 2 in 2024.

CONCLUSIONS

Our findings will yield insights into improving medication adherence and C-MCIs among American Indian adults, consistent with CNO's State of the Nation's Health Report 2017 goal of reducing T2D and its complications.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

PRR1-10.2196/39193.

A Machine-Learning Based Approach for Predicting Older Adults' Adherence to Technology-Based Cognitive Training

Adequate adherence is a necessary condition for success with any intervention, including for computerized cognitive training designed to mitigate age-related cognitive decline. Tailored prompting systems offer promise for promoting adherence and facilitating intervention success. However, developing adherence support systems capable of just-in-time adaptive reminders requires understanding the factors that predict adherence, particularly an imminent adherence lapse. In this study we built machine learning models to predict participants' adherence at different levels (overall and weekly) using data collected from a previous cognitive training intervention. We then built machine learning models to predict adherence using a variety of baseline measures (demographic, attitudinal, and cognitive ability variables), as well as deep learning models to predict the next week's adherence using variables derived from training interactions in the previous week. Logistic regression models with selected baseline variables were able to predict overall adherence with moderate accuracy (AUROC: 0.71), while some recurrent neural network models were able to predict weekly adherence with high accuracy (AUROC: 0.84-0.86) based on daily interactions. Analysis of the post hoc explanation of machine learning models revealed that general self-efficacy, objective memory measures, and technology self-efficacy were most predictive of participants' overall adherence, while time of training, sessions played, and game outcomes were predictive of the next week's adherence. Machine-learning based approaches revealed that both individual difference characteristics and previous intervention interactions provide useful information for predicting adherence, and these insights can provide initial clues as to who to target with adherence support strategies and when to provide support. This information will inform the development of a technology-based, just-in-time adherence support systems.

SANE (Easy Gait Analysis System): Towards an AI-Assisted Automatic Gait-Analysis

The gait cycle of humans may be influenced by a range of variables, including neurological, orthopedic, and pathological conditions. Thus, gait analysis has a broad variety of applications, including the diagnosis of neurological disorders, the study of disease development, the assessment of the efficacy of a treatment, postural correction, and the evaluation and enhancement of sport performances. While the introduction of new technologies has resulted in substantial advancements, these systems continue to struggle to achieve a right balance between cost, analytical accuracy, speed, and convenience. The target is to provide low-cost support to those with motor impairments in order to improve their quality of life. The article provides a novel automated approach for motion characterization that makes use of artificial intelligence to perform real-time analysis, complete automation, and non-invasive, markerless analysis. This automated procedure enables rapid diagnosis and prevents human mistakes. The gait metrics obtained by the two motion tracking systems were compared to show the effectiveness of the proposed methodology.

Use of machine learning to identify patients at risk of sub-optimal adherence: study based on real-world data from 10,929 children using a connected auto-injector device

BACKGROUND

Our aim was to develop a machine learning model, using real-world data captured from a connected auto-injector device and from early indicators from the first 3 months of treatment, to predict sub-optimal adherence to recombinant human growth hormone (r-hGH) in patients with growth disorders.

METHODS

Adherence to r-hGH treatment was assessed in children (aged < 18 years) who started using a connected auto-injector device (easypod™), and transmitted injection data for ≥ 12 months. Adherence in the following 3, 6, or 9 months after treatment start was categorized as optimal (≥ 85%) versus sub-optimal (< 85%). Logistic regression and tree-based models were applied.

RESULTS

Data from 10,929 children showed that a random forest model with mean and standard deviation of adherence over the first 3 months, infrequent transmission of data, not changing certain comfort settings, and starting treatment at an older age was important in predicting the risk of sub-optimal adherence in the following 3, 6, or 9 months. Sensitivities ranged between 0.72 and 0.77, and specificities between 0.80 and 0.81.

CONCLUSIONS

To the authors' knowledge, this is the first attempt to integrate a machine learning model into a digital health ecosystem to help healthcare providers to identify patients at risk of sub-optimal adherence to r-hGH in the following 3, 6, or 9 months. This information, together with patient-specific indicators of sub-optimal adherence, can be used to provide support to at-risk patients and their caregivers to achieve optimal adherence and, subsequently, improve clinical outcomes.

Adherence to and persistence with antidiabetic medications and associations with clinical and economic outcomes in people with type 2 diabetes mellitus: A systematic literature review

We designed a systematic literature review to identify available evidence on adherence to and persistence with antidiabetic medication in people with type 2 diabetes (T2D). Electronic screening and congress searches identified real-world noninterventional studies (published between 2010 and October 2020) reporting estimates of adherence to and persistence with antidiabetic medication in adults with T2D, and associations with glycaemic control, microvascular and/or macrovascular complications, hospitalizations and healthcare costs. Ninety-two relevant studies were identified, the majority of which were retrospective and reported US data. The proportions of patients considered adherent (median [range] 51.2% [9.4%-84.3%]) or persistent (median [range] 47.7% [16.9%-94.0%]) varied widely across studies. Multiple studies reported an association between greater adherence/persistence and greater reductions in glycated haemoglobin levels. Better adherence/persistence was associated with fewer microvascular and/or macrovascular outcomes, although there was little consistency across studies in terms of which outcomes were improved. More adherent and more persistent patients were typically less likely to be hospitalized or to have emergency department visits/admissions and spent fewer days in hospital annually than less adherent/persistent patients. Greater adherence and persistence were generally associated with lower hospitalization costs, higher pharmacy costs and lower or budget-neutral total healthcare costs compared with lower adherence/persistence. In conclusion, better adherence and persistence in people with T2D is associated with lower rates of microvascular and/or macrovascular outcomes and inpatient hospitalization, and lower or budget-neutral total healthcare expenditure. Education and treatment strategies to address suboptimal adherence and persistence are needed to improve clinical and economic outcomes.

Predicting Parkinson’s Disease Progression: Evaluation of Ensemble Methods in Machine Learning

Parkinson’s disease (PD) is a complex neurodegenerative disease. Accurate diagnosis of this disease in the early stages is crucial for its initial treatment. This paper aims to present a comparative study on the methods developed by machine learning techniques in PD diagnosis. We rely on clustering and prediction learning approaches to perform the comparative study. Specifically, we use different clustering techniques for PD data clustering and support vector regression ensembles to predict Motor-UPDRS and Total-UPDRS. The results are then compared with the other prediction learning approaches, multiple linear regression, neurofuzzy, and support vector regression techniques. The comparative study is performed on a real-world PD dataset. The prediction results of data analysis on a PD real-world dataset revealed that expectation-maximization with the aid of SVR ensembles can provide better prediction accuracy in relation to decision trees, deep belief network, neurofuzzy, and support vector regression combined with other clustering techniques in the prediction of Motor-UPDRS and Total-UPDRS.

Determining the optimal threshold for medication adherence in adult asthma patients: an analysis of British Columbia administrative health database in Canada

OBJECTIVE

This study investigated the association between varying cutoffs for Medication Adherence (MA) among physician-diagnosed asthma patients and subsequent association with asthma exacerbation.

METHODS

We linked four administrative health databases obtained from the Population Data in British Columbia. Index cases were physician-diagnosed asthma patients between January 1, 1998, to December 31, 1999, aged 18 years and older. Patients were prospectively assessed in the follow-up period from January 1, 2000, to December 31, 2018, to identify asthma exacerbation. Two proxy measures were used to assess MA: the proportion of days covered (PDC) and the medication possession ratio (MPR). Using the generalized estimating equation (GEE) logistic regression adjusted for patient covariates, the outcome of "asthma exacerbation" was modeled against varying MA cutoffs; excellent '≥0.90'; very good '0.80-0.89'; good '0.70-0.799'; moderate '0.6-0.699'; mild '0.50-0.599' compared to poor '<0.50' for both PDC and MPR.

RESULTS

The sample included 68,211 physician-diagnosed asthma patients with a mean age of 48.2 years and 59.3% females. The adjusted odds ratios (OR) and 95% confidence interval (CI) at the various cutoff for PDC-levels predicting asthma exacerbation events were: Excellent MA [OR = 0.84, 95% (0.82-0.86), very good MA [OR: 0.86, (0.83, 0.89), good MA [0.91, (0.88-0.94)]; moderate MA [0.93, (0.90-0.96)]; mild MA [0.95, (0.92-0.98)]; compared to poor MA level. Threshold levels for both the PDC and MPR measure greater than 0.80 provided optimal threshold associated with over 15% reduced likelihood of experiencing asthma exacerbations.

CONCLUSION

Intervention aimed at improving asthma exacerbation events in adult asthma patients should encourage increased medication adherence threshold level greater than 0.80.

Supplemental data for this article is available online at at www.tandfonline.com/ijas .

Artificial Intelligence Solutions to Increase Medication Adherence in Patients With Non-communicable Diseases

Artificial intelligence (AI) tools are increasingly being used within healthcare for various purposes, including helping patients to adhere to drug regimens. The aim of this narrative review was to describe: (1) studies on AI tools that can be used to measure and increase medication adherence in patients with non-communicable diseases (NCDs); (2) the benefits of using AI for these purposes; (3) challenges of the use of AI in healthcare; and (4) priorities for future research. We discuss the current AI technologies, including mobile phone applications, reminder systems, tools for patient empowerment, instruments that can be used in integrated care, and machine learning. The use of AI may be key to understanding the complex interplay of factors that underly medication non-adherence in NCD patients. AI-assisted interventions aiming to improve communication between patients and physicians, monitor drug consumption, empower patients, and ultimately, increase adherence levels may lead to better clinical outcomes and increase the quality of life of NCD patients. However, the use of AI in healthcare is challenged by numerous factors; the characteristics of users can impact the effectiveness of an AI tool, which may lead to further inequalities in healthcare, and there may be concerns that it could depersonalize medicine. The success and widespread use of AI technologies will depend on data storage capacity, processing power, and other infrastructure capacities within healthcare systems. Research is needed to evaluate the effectiveness of AI solutions in different patient groups and establish the barriers to widespread adoption, especially in light of the COVID-19 pandemic, which has led to a rapid increase in the use and development of digital health technologies.

Self-reported medication adherence measurement tools: Some options to avoid a legal minefield

WHAT IS KNOWN AND OBJECTIVE

Self-report questionnaires are used to measure medication adherence, often times both clinically and for research purposes. Despite the presence of several published tools, some may have prohibitive licensure and fee requirements, which researchers should be aware of prior to using them. This paper presents a summary of selected self-report measures, which have been developed and validated in various health conditions and can be used free of cost.

COMMENT

Our review identified self-report tools that are valid to measure medication adherence in different chronic health conditions. Most of these tools measure both intentional and unintentional non-adherence and have shown good correlation with relevant clinical outcomes.

WHAT IS NEW AND CONCLUSION

Given the potential copyright risks associated with using some of the self-report measures of adherence, an improved awareness and understanding of the available self-report questionnaires will better facilitate the decision by researchers to select appropriate tools relevant to their studies.

Measuring the multiple domains of medication nonadherence: findings from a Delphi survey of adherence experts

Consensus on a gold-standard measure of patient medication nonadherence has been elusive, in part because medication nonadherence involves multiple, distinct behaviors across three phases (initiation, implementation, and persistence). To assess these behaviors, multiple measurement approaches may be needed. The purpose of this study was to identify expert-recommended approaches to measuring nonadherence behaviors. Thirty medication nonadherence experts were e-mailed two consecutive surveys. In both, respondents rated their agreement with definitions of nonadherence behaviors and measurement approaches. In the second survey, respondents rated the suitability of each measurement approach for assessing each behavior and identified the optimal measurement approach for each behavior. Consensus was achieved for eight patient medication nonadherence behaviors: not filling initial prescription and not taking first dose (noninitiation); refilling prescription late, missing doses, taking extra doses, taking doses at wrong time, and improperly administering medication (incorrect implementation); and discontinuing medication early (nonpersistence). Consensus was achieved for seven measurement approaches: self-report, prescription fill data, pill count, drug levels, electronic drug monitoring (EDM), smart technology, and direct observation. Self-report questionnaires were most commonly rated "at least somewhat suitable" for measuring behaviors. EDM was rated as optimal for measuring missing doses, taking extra doses, and taking doses at the wrong time. Prescription fill data were rated as optimal for not filling initial prescription, refilling late, and discontinuing. Direct observation was rated as optimal for measuring improper administration. Suitable and optimal measurement approaches varied across nonadherence behaviors. Researchers should select the measurement approach best suited to assessing the behavior(s) targeted in their research.

Using Machine Learning to Predict Early Preparation of Pharmacy Prescriptions at PSMMC - a Comparison of Four Machine Learning Algorithms

BACKGROUND

Patient satisfaction is one of the primary Key Performance Indicator (KPI) goal of health care service, and it creates many reasons for implementing research, plans, and innovations to achieve it for a better quality of life. Cutting Patient waiting time would increase patient satisfaction.

OBJECTIVE

A healthcare framework has been constructed utilizing a machine learning approach to construct an early predicting preparation model of pharmacy prescriptions and the worthiness of changing the outpatient pharmacy workflow.

METHODS

Data sets were retrieved between Januarys and June 2019 from Prince Sultan Military Medical City, Riyadh, KSA, for all patients who visited the clinics or discharged with pharmacy prescriptions. Included (1048575) instances and composed of (11) attributes. The evaluation criteria to compare the four algorithms were based on precision, Recall, True Positive Rate, False Negative Rate, F-measure, and Area under the curve.

RESULTS

Overall, 94.88% of patient's shows at the pharmacy, female represents 58.89% of the data set while male represents 41.1%. RT gives the highest accuracy, with 97.22% in comparison to the other algorithms.

CONCLUSION

The suggestion to change the pharmacy workflow is worth increasing patient satisfaction and overall the quality of the care.

Optimal cut-off points for adherence measure among patients with type 2 diabetes in primary care clinics: a retrospective analysis

Background

Medication adherence measures are often dichotomized to classify patients into those with good or poor adherence using a cut-off value ⩾80%, but this cut-off may not be universal across diseases or medication classes. This study aimed to examine the cut-off value that optimally distinguish good and poor adherence by using the medication possession ratio (MPR) and proportion of days covered (PDC) as adherence measures and glycated hemoglobin (HbA1c) as outcome measure among type 2 diabetes mellitus (T2DM) patients.

Method

We used pharmacy dispensing data of 1461 eligible T2DM patients from public primary care clinics in Malaysia treated with oral antidiabetic drugs between January 2018 and May 2019. Adherence rates were calculated during the period preceding the HbA1c measurement. Adherence cut-off values for the following conditions were compared: adherence measure (MPR versus PDC), assessment period (90-day versus 180-day), and HbA1c target (⩽7.0% versus ⩽8.0%).

Results

The optimal adherence cut-offs for MPR and PDC in predicting HbA1c ⩽7.0% ranged between 86.1% and 98.3% across the two assessment periods. In predicting HbA1c ⩽8.0%, the optimal adherence cut-offs ranged from 86.1% to 92.8%. The cut-off value was notably higher with PDC as the adherence measure, shorter assessment period, and a stricter HbA1c target (⩽7.0%) as outcome.

Conclusion

We found that optimal adherence cut-off appeared to be slightly higher than the conventional value of 80%. The adherence thresholds may vary depending on the length of assessment period and outcome definition but a reasonably wise cut-off to distinguish good versus poor medication adherence to be clinically meaningful should be at 90%.

The Need for Ethnoracial Equity in Artificial Intelligence for Diabetes Management: Review and Recommendations

There is clear evidence to suggest that diabetes does not affect all populations equally. Among adults living with diabetes, those from ethnoracial minority communities—foreign-born, immigrant, refugee, and culturally marginalized—are at increased risk of poor health outcomes. Artificial intelligence (AI) is actively being researched as a means of improving diabetes management and care; however, several factors may predispose AI to ethnoracial bias. To better understand whether diabetes AI interventions are being designed in an ethnoracially equitable manner, we conducted a secondary analysis of 141 articles included in a 2018 review by Contreras and Vehi entitled “Artificial Intelligence for Diabetes Management and Decision Support: Literature Review.” Two members of our research team independently reviewed each article and selected those reporting ethnoracial data for further analysis. Only 10 articles (7.1%) were ultimately selected for secondary analysis in our case study. Of the 131 excluded articles, 118 (90.1%) failed to mention participants’ ethnic or racial backgrounds. The included articles reported ethnoracial data under various categories, including race (n=6), ethnicity (n=2), race/ethnicity (n=3), and percentage of Caucasian participants (n=1). Among articles specifically reporting race, the average distribution was 69.5% White, 17.1% Black, and 3.7% Asian. Only 2 articles reported inclusion of Native American participants. Given the clear ethnic and racial differences in diabetes biomarkers, prevalence, and outcomes, the inclusion of ethnoracial training data is likely to improve the accuracy of predictive models. Such considerations are imperative in AI-based tools, which are predisposed to negative biases due to their black-box nature and proneness to distributional shift. Based on our findings, we propose a short questionnaire to assess ethnoracial equity in research describing AI-based diabetes interventions. At this unprecedented time in history, AI can either mitigate or exacerbate disparities in health care. Future accounts of the infancy of diabetes AI must reflect our early and decisive action to confront ethnoracial inequities before they are coded into our systems and perpetuate the very biases we aim to eliminate. If we take deliberate and meaningful steps now toward training our algorithms to be ethnoracially inclusive, we can architect innovations in diabetes care that are bound by the diverse fabric of our society.

Machine learning in liver transplantation: a tool for some unsolved questions?

Machine learning has recently been proposed as a useful tool in many fields of Medicine, with the aim of increasing diagnostic and prognostic accuracy. Models based on machine learning have been introduced in the setting of solid organ transplantation too, where prognosis depends on a complex, multidimensional and nonlinear relationship between variables pertaining to the donor, the recipient and the surgical procedure. In the setting of liver transplantation, machine learning models have been developed to predict pretransplant survival in patients with cirrhosis, to assess the best donor-to-recipient match during allocation processes, and to foresee postoperative complications and outcomes. This is a narrative review on the role of machine learning in the field of liver transplantation, highlighting strengths and pitfalls, and future perspectives.

A Clinical Prediction Model of Medication Adherence in Hypertensive Patients in a Chinese Community Hospital in Beijing

BACKGROUND

Hypertension remains a global health problem. Since, there is a significant positive correlation between antihypertensive medication adherence and blood pressure control, it is therefore of great importance to elucidate the determinants of adherence to antihypertensive medications among hypertensive patients.

METHODS

Hereby, we retrospectively analyzed the medical records of a hypertensive cohort recruited from a community hospital in Beijing, China, to investigate the factors affecting adherence to antihypertensive medications using decision trees. In addition, all data were assigned into a training set (75%) and testing set (25%) by the random number seed method to build and validate a compliance predictive model. We identified that how many times patients became nonadherent to antihypertensive medications in the year before the first prescription, types of antihypertensive drugs used in the year before the first prescription, body weight, smoking history, total number of hospital visits in the past year, total number of days of medication use in the year before enrollment, age, total number of outpatient follow-ups in the year after the first prescription, and concurrent diabetes greatly affected the compliance to antihypertensive medications.

RESULTS

The compliance predictive model we built showed a 0.78 sensitivity and 0.69 specificity for the prediction of the compliance to antihypertensive medications, with an area under the representative operating characteristics curve of 0.810.

CONCLUSIONS

Our data provide new insights into the improvements of the compliance to antihypertensive medications, which is beneficial for the management of hypertension, and the compliance predictive model may be used in community-based hypertension management.

Association of Self-Reported Medication Adherence with Potentially Inappropriate Medications in Elderly Patients: A Cross-Sectional Pilot Study

BACKGROUND

Polypharmacy (PP) and potentially inappropriate medications (PIMs) cause problematic drug-related issues in elderly patients; however, little is known about the association between medication adherence and PP and PIMs. This study evaluated the association of self-reported medication adherence with PP and PIMs in elderly patients.

METHODS

A cross-sectional pilot study was conducted using data collected from electronic medical records of 142 self-administering patients aged ≥65 years, excluding emergency hospitalization cases. Self-reported medication adherence was assessed using the visual analogue scale (VAS).

RESULTS

Of the 142 patients, 91 (64.1%) had PP and 80 (56.3%) used at least one PIM. In univariate analysis, patients with a VAS score of 100% had a significantly higher number of female patients and ≥1 PIM use compared to other patients. We found no association between the VAS score and PP. In multivariable analysis, the use of PIMs was significantly associated with a VAS score of 100% (odds ratio = 2.32; 95% confidence interval = 1.16-4.72; p = 0.017).

CONCLUSIONS

Use of PIMs by elderly patients is significantly associated with self-reported medication adherence. Pharmacists should pay more attention to prescribed medications of self-administering elderly patients in order to improve their prescribing quality.

Artificial Intelligence: The Future for Diabetes Care

Artificial intelligence (AI) is a fast-growing field and its applications to diabetes, a global pandemic, can reform the approach to diagnosis and management of this chronic condition. Principles of machine learning have been used to build algorithms to support predictive models for the risk of developing diabetes or its consequent complications. Digital therapeutics have proven to be an established intervention for lifestyle therapy in the management of diabetes. Patients are increasingly being empowered for self-management of diabetes, and both patients and health care professionals are benefitting from clinical decision support. AI allows a continuous and burden-free remote monitoring of the patient's symptoms and biomarkers. Further, social media and online communities enhance patient engagement in diabetes care. Technical advances have helped to optimize resource use in diabetes. Together, these intelligent technical reforms have produced better glycemic control with reductions in fasting and postprandial glucose levels, glucose excursions, and glycosylated hemoglobin. AI will introduce a paradigm shift in diabetes care from conventional management strategies to building targeted data-driven precision care.

Artificial Intelligence in Pharmacoepidemiology: A Systematic Review. Part 1-Overview of Knowledge Discovery Techniques in Artificial Intelligence

Aim

To perform a systematic review on the application of artificial intelligence (AI) based knowledge discovery techniques in pharmacoepidemiology.

Study Eligibility Criteria

Clinical trials, meta-analyses, narrative/systematic review, and observational studies using (or mentioning articles using) artificial intelligence techniques were eligible. Articles without a full text available in the English language were excluded.

Data Sources

Articles recorded from 1950/01/01 to 2019/05/06 in Ovid MEDLINE were screened.

Participants

Studies including humans (real or simulated) exposed to a drug.

Results

In total, 72 original articles and 5 reviews were identified via Ovid MEDLINE. Twenty different knowledge discovery methods were identified, mainly from the area of machine learning (66/72; 91.7%). Classification/regression (44/72; 61.1%), classification/regression + model optimization (13/72; 18.0%), and classification/regression + features selection (12/72; 16.7%) were the three most frequent tasks in reviewed literature that machine learning methods has been applied to solve. The top three used techniques were artificial neural networks, random forest, and support vector machines models.

Conclusions

The use of knowledge discovery techniques of artificial intelligence techniques has increased exponentially over the years covering numerous sub-topics of pharmacoepidemiology.

Systematic Review Registration

Systematic review registration number in PROSPERO: CRD42019136552.

Restructured society and environment: A review on potential technological strategies to control the COVID-19 pandemic

The emergence of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) in China at December 2019 had led to a global outbreak of coronavirus disease 2019 (COVID-19) and the disease started to spread all over the world and became an international public health issue. The entire humanity has to fight in this war against the unexpected and each and every individual role is important. Healthcare system is doing exceptional work and the government is taking various measures that help the society to control the spread. Public, on the other hand, coordinates with the policies and act accordingly in most state of affairs. But the role of technologies in assisting different social bodies to fight against the pandemic remains hidden. The intention of our study is to uncover the hidden roles of technologies that ultimately help for controlling the pandemic. On investigating, it is found that the strategies utilizing potential technologies would yield better benefits and these technological strategies can be framed either to control the pandemic or to support the confinement of the society during pandemic which in turn aids in controlling the spreading of infection. This study enlightens the various implemented technologies that assists the healthcare systems, government and public in diverse aspects for fighting against COVID-19. Furthermore, the technological swift that happened during the pandemic and their influence in the environment and society is discussed. Besides the implemented technologies, this work also deals with untapped potential technologies that have prospective applications in controlling the pandemic circumstances. Alongside the various discussion, our suggested solution for certain situational issues is also presented.

Artificial Intelligence and Big Data in Diabetes Care: A Position Statement of the Italian Association of Medical Diabetologists

Since the last decade, most of our daily activities have become digital. Digital health takes into account the ever-increasing synergy between advanced medical technologies, innovation, and digital communication. Thanks to machine learning, we are not limited anymore to a descriptive analysis of the data, as we can obtain greater value by identifying and predicting patterns resulting from inductive reasoning. Machine learning software programs that disclose the reasoning behind a prediction allow for “what-if” models by which it is possible to understand if and how, by changing certain factors, one may improve the outcomes, thereby identifying the optimal behavior. Currently, diabetes care is facing several challenges: the decreasing number of diabetologists, the increasing number of patients, the reduced time allowed for medical visits, the growing complexity of the disease both from the standpoints of clinical and patient care, the difficulty of achieving the relevant clinical targets, the growing burden of disease management for both the health care professional and the patient, and the health care accessibility and sustainability. In this context, new digital technologies and the use of artificial intelligence are certainly a great opportunity. Herein, we report the results of a careful analysis of the current literature and represent the vision of the Italian Association of Medical Diabetologists (AMD) on this controversial topic that, if well used, may be the key for a great scientific innovation. AMD believes that the use of artificial intelligence will enable the conversion of data (descriptive) into knowledge of the factors that “affect” the behavior and correlations (predictive), thereby identifying the key aspects that may establish an improvement of the expected results (prescriptive). Artificial intelligence can therefore become a tool of great technical support to help diabetologists become fully responsible of the individual patient, thereby assuring customized and precise medicine. This, in turn, will allow for comprehensive therapies to be built in accordance with the evidence criteria that should always be the ground for any therapeutic choice.

Random forest machine learning algorithm predicts virologic outcomes among HIV infected adults in Lausanne, Switzerland using electronically monitored combined antiretroviral treatment adherence

Machine Learning (ML) can improve the analysis of complex and interrelated factors that place adherent people at risk of viral rebound. Our aim was to build ML model to predict RNA viral rebound from medication adherence and clinical data. Patients were followed up at the Swiss interprofessional medication adherence program (IMAP). Sociodemographic and clinical variables were retrieved from the Swiss HIV Cohort Study (SHCS). Daily electronic medication adherence between 2008-2016 were analyzed retrospectively. Predictor variables included: RNA viral load (VL), CD4 count, duration of ART, and adherence. Random Forest, was used with 10 fold cross validation to predict the RNA class for each data observation. Classification accuracy metrics were calculated for each of the 10-fold cross validation holdout datasets. The values for each range from 0 to 1 (better accuracy). 383 HIV+ patients, 56% male, 52% white, median (Q1, Q3): age 43 (36, 50), duration of electronic monitoring of adherence 564 (200, 1333) days, CD4 count 406 (209, 533) cells/mm3, time since HIV diagnosis was 8.4 (4, 13.5) years, were included. Average model classification accuracy metrics (AUC and F1) for RNA VL were 0.6465 and 0.7772, respectively. In conclusion, combining adherence with other clinical predictors improve predictions of RNA.

Adherence to Anticoagulation and Risk of Stroke Among Medicare Beneficiaries Newly Diagnosed with Atrial Fibrillation

INTRODUCTION

The objective of this study was to compare the risk of stroke in atrial fibrillation (AF) with adherent use of oral anticoagulation (OAC), non-adherent use, and non-use of OAC.

METHODS

Using 2013-2016 Medicare claims data, we identified patients newly diagnosed with AF in 2014-2015 and collected prescriptions filled for OAC in the 12 months after AF diagnosis (n = 39,272). We categorized participants each day into three time-dependent exposures: adherent use (≥ 80% of the previous 30 days covered with OAC), non-adherent use (0-80% covered with OAC), and non-use (0%). We constructed Cox proportional hazards models to estimate the association between time-dependent exposures and time to stroke, adjusting for demographics and clinical characteristics.

RESULTS

The sample included 39,272 patients. Study participants spent 35.0% of the follow-up period in the adherent use exposure category, 10.9% in the non-adherent category, and 54.0% in the non-use category. OAC adherent use [hazard ratio (HR) 0.62; 95% confidence interval (CI) 0.52-0.74] and non-adherent use (HR 0.74; 95% CI 0.57-0.95) were associated with lower hazards of stroke than non-use. Adherent use of DOAC (HR 0.54; 95% CI 0.42-0.69) and warfarin (HR 0.70; 95% CI 0.56-0.89) was associated with lower risk of stroke than non-use, but the risk of stroke did not statistically differ between DOAC and warfarin adherent use (HR 0.77; 95% CI 0.56-1.04).

DISCUSSION

Although adherence to OAC reduces stroke risk by nearly 40%, newly diagnosed AF patients in Medicare adhere to OAC on average only one third of the first year after AF diagnosis.

Refill Adherence Measures and Its Association with Economic, Clinical, and Humanistic Outcomes Among Pediatric Patients: A Systematic Review

Although refill adherence measures (RAMs) are widely reviewed on their use among adult patients, existing reviews on adherence among children have only focused on self-report measures and electronic monitoring. Hence, this systematic review aims to examine the use of RAMs and their association with economic, clinical, and humanistic outcomes (ECHO) among pediatric patients. A literature search was conducted in Pubmed, Embase, CINAHL, and PsycINFO. Studies published in English involving subjects aged ≤18 years were included if RAMs were analyzed with ECHO. Of the 35 included studies, the majority (n = 33) were conducted in high-income countries. Asthma was the most common condition (n = 9) studied. Overall, 60.6% of 33 clinical outcomes reported among 22 studies was positive (improved clinical outcomes with improved adherence), while 21.9% of 32 economic outcomes reported among 16 studies was positive (reduced healthcare utilization or cost outcomes with improved adherence). Only four studies evaluated the relationship of adherence with 11 humanistic outcomes, where the majority (72.7%) were considered unclear. RAMs are associated with ECHO and can be considered for use in the pediatric population. Future studies could explore the use of RAMs in low-income countries, and the association of RAMs with quality of life.

Predictive models of medication non-adherence risks of patients with T2D based on multiple machine learning algorithms

OBJECTIVE

Medication adherence plays a key role in type 2 diabetes (T2D) care. Identifying patients with high risks of non-compliance helps individualized management, especially for China, where medical resources are relatively insufficient. However, models with good predictive capabilities have not been studied. This study aims to assess multiple machine learning algorithms and screen out a model that can be used to predict patients' non-adherence risks.

METHODS

A real-world registration study was conducted at Sichuan Provincial People's Hospital from 1 April 2018 to 30 March 2019. Data of patients with T2D on demographics, disease and treatment, diet and exercise, mental status, and treatment adherence were obtained by face-to-face questionnaires. The medication possession ratio was used to evaluate patients' medication adherence status. Fourteen machine learning algorithms were applied for modeling, including Bayesian network, Neural Net, support vector machine, and so on, and balanced sampling, data imputation, binning, and methods of feature selection were evaluated by the area under the receiver operating characteristic curve (AUC). We use two-way cross-validation to ensure the accuracy of model evaluation, and we performed a posteriori test on the sample size based on the trend of AUC as the sample size increase.

RESULTS

A total of 401 patients out of 630 candidates were investigated, of which 85 were evaluated as poor adherence (21.20%). A total of 16 variables were selected as potential variables for modeling, and 300 models were built based on 30 machine learning algorithms. Among these algorithms, the AUC of the best capable one was 0.866±0.082. Imputing, oversampling and larger sample size will help improve predictive ability.

CONCLUSIONS

An accurate and sensitive adherence prediction model based on real-world registration data was established after evaluating data filling, balanced sampling, and so on, which may provide a technical tool for individualized diabetes care.

Big Data Analytics and Sensor-Enhanced Activity Management to Improve Effectiveness and Efficiency of Outpatient Medical Rehabilitation

Numerous societal trends are compelling a transition from inpatient to outpatient venues of care for medical rehabilitation. While there are advantages to outpatient rehabilitation (e.g., lower cost, more relevant to home and community function), there are also challenges including lack of information about how patient progress observed in the outpatient clinic translates into improved functional performance at home. At present, outpatient providers must rely on patient-reported information about functional progress (or lack thereof) at home and in the community. Information and communication technologies (ICT) offer another option—data collected about the patient’s adherence, performance and progress made on home exercises could be used to help guide course corrections between clinic visits, enhancing effectiveness and efficiency of outpatient care. In this article, we describe our efforts to explore use of sensor-enhanced home exercise and big data analytics in medical rehabilitation. The goal of this work is to demonstrate how sensor-enhanced exercise can improve rehabilitation outcomes for patients with significant neurological impairment (e.g., from stroke, traumatic brain injury, and spinal cord injury). We provide an overview of big data analysis and explain how it may be used to optimize outpatient rehabilitation, creating a more efficient model of care. We describe our planned development efforts to build advanced analytic tools to guide home-based rehabilitation and our proposed randomized trial to evaluate effectiveness and implementation of this approach.

Association of Potentially Modifiable Diabetes Care Factors With Glycemic Control in Patients With Insulin-Treated Type 2 Diabetes

IMPORTANCE

Numerous factors are associated with the ability of patients with type 2 diabetes to achieve optimal glycemic control. However, many of these factors are not modifiable by quality improvement interventions. In contrast, the structure of how diabetes care is delivered, such as whether patients visit an endocrinologist or how prescriptions are filled, is potentially modifiable, yet its associations with glycemic control have not been rigorously evaluated.

OBJECTIVE

To investigate the association of diabetes care delivery with glycemic control in patients with type 2 diabetes using insulin.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective cohort study used baseline claims and laboratory insurer data within a large pragmatic trial to identify individuals with type 2 diabetes using insulin with data for at least 1 hemoglobin A1c (HbA1c) test result from before trial randomization (July 1, 2014, to October 5, 2016) and for key nonmodifiable patient factors as well as diabetes care delivery and behavioral factors measured before the HbA1c test. Analyses were conducted from February 4, 2017, to November 13, 2018.

MAIN OUTCOMES AND MEASURES

Multivariable modified Poisson regression was used to evaluate the independent associations of nonmodifiable patient factors and potentially modifiable diabetes care delivery and patient behavioral factors with achieving adequate diabetes control (ie, HbA1c level <8%). The extent of measured variation explained in glycemic control by these factors was also explored using pseudo R2 and C statistics.

RESULTS

Of 1423 patients included, 565 (39.7%) were women, and the mean (SD) age was 56.4 (9.0) years. In total, 690 (48.5%) had HbA1c levels less than 8%. Age (relative risk [RR] per 1-unit increase, 1.01; 95% CI, 1.00-1.02), persistent use of basal insulin (RR, 1.20; 95% CI, 1.00-1.43), more frequent filling of glucose self-testing supplies (RR, 1.01; 95% CI, 1.01-1.02), visiting an endocrinologist (RR, 1.41; 95% CI, 1.19-1.67), and receipt of insulin prescriptions by mail order (RR, 1.23; 95% CI, 1.03-1.48) were all independently associated with adequate control. Measured potentially modifiable diabetes care factors explained more variation in adequate glycemic control than measured nonmodifiable patient factors (C statistic, 0.661 vs 0.598; pseudo R2 = 0.11 vs 0.04).

CONCLUSIONS AND RELEVANCE

These findings suggest that for patients with type 2 diabetes using insulin, the way in which care is delivered may be more strongly associated with achieving adequate control of HbA1c levels than patient factors that cannot be altered are. Given the potential for intervention, these care delivery factors could be the focus of efforts to improve diabetes outcomes.

Optimal threshold of adherence to lipid lowering drugs in predicting acute coronary syndrome, stroke, or mortality: A cohort study

Objective

Thresholds defining medication adherence are rarely evidence-based. A threshold of 0.8 is typically presumed to achieve improved outcomes. We aimed to assess the optimal threshold of adherence to lipid-lowering drugs (LLD) in predicting cardiovascular-related (CV) outcomes in patients with hypertension.

Design

Cohort study of new users of LLDs.

Setting

Comprehensive healthcare administrative databases of the province of Alberta (Canada) from 2008 to 2016.

Participants

Patients with hypertension, who were new users of LLDs. Patients who had the outcomes prior to the initiation of LLD were excluded.

Main outcomes measures

Hospitalization for acute coronary syndrome (ACS)/stroke, CV-related mortality and all-cause mortality.

Statistical analysis

Adherence to LLDs was assessed as the proportion of days covered (PDC) by any LLD, from drug initiation to censoring, outcome, or study end. Three methods were used to assess the threshold: Contal and O'Quigley method, minimum distance method, and Youden's J index. Cox regressions were used to assess the risk associated with each method-specific threshold and Akaike information criteria were used to retain the optimal threshold after adjustment.

Results

52229 patients were included; 4.0% were hospitalized for ACS/stroke, 3.4% died, and 1.3% died from CV-related cause. In predicting ACS/stroke, CV-related and all-cause mortality, the optimal adherence threshold was 0.52 (range: 0.51–0.54), 0.79 (0.45–0.87), and 0.84 (0.79–0.89), respectively. These results were consistent among patients aged ≥ 65 years (n = 19804). However, the results varied among those aged < 65 years, where the incidence rates of outcomes were low.

Conclusion

In new-users of LLDs with hypertension, approximately 50% days covered by LLDs may be enough to prevent long-term occurrence of ACS, or stroke. However, a threshold near 0.80 may be needed to prevent or reduce the risk of all-cause or CV-related mortality.

Effectiveness of Targeted Insulin-Adherence Interventions for Glycemic Control Using Predictive Analytics Among Patients With Type 2 Diabetes: A Randomized Clinical Trial

Importance

Patient adherence to antidiabetic medications, especially insulin, remains poor, leading to adverse outcomes and increased costs. Most adherence interventions have only been modestly effective, partly because they are not targeted to patients who could benefit most.

Objective

To evaluate whether delivering more intensive insulin-adherence interventions only to individuals with type 2 diabetes predicted to benefit most was more effective than delivering a lower-intensity intervention to a larger group of unselected individuals.

Design, Setting, and Participants

This 3-arm pragmatic randomized clinical trial used data from Horizon, the largest health insurer in New Jersey, on 6000 participants 18 years or older with type 2 diabetes who were receiving basal insulin. Patients were excluded if they were insured by Medicaid or Medicare or had fewer than 3 months of continuous enrollment. The study was conducted from July 7, 2016, through October 5, 2017. Analyses were conducted from February 5 to September 24, 2018.

Interventions

Eligible patients were randomized to 3 arms in a 1:1:1 ratio. Randomization was stratified based on baseline availability of 1 or more glycated hemoglobin A1c (HbA1c) test values. All arms were designed to cost the same, and each cohort received a tailored pharmacist telephone consultation varying based on (1) proportion receiving the intervention and (2) intensity, including follow-up frequency and cointerventions. Arm 1 offered a low-intensity intervention to all patients. Arm 2 offered a moderate-intensity intervention to 60% of patients based on their predicted risk of insulin nonadherence. Arm 3 offered a high-intensity intervention to 40% of patients based on glycemic control and predicted risk of insulin nonadherence.

Main Outcomes and Measures

The primary outcome was insulin persistence. Secondary outcomes were changes in HbA1c level and health care utilization. Outcomes were evaluated in arms 2 and 3 vs arm 1 using claims data, intention-to-treat principles, and multiple imputation for missing values in the 12-month follow-up.

Results

Among 6000 participants, mean (SD) age was 55.9 (11.0) years and 3344 (59.8%) were male. Compared with arm 1, insulin nonpersistence did not differ in arm 2 (relative risk, 0.88; 95% CI, 0.75-1.03) or arm 3 (relative risk, 0.91; 95% CI, 0.77-1.06). Glycemic control was similar in arm 2 and arm 1 (absolute HbA1c level difference, -0.15%; 95% CI, -0.34% to 0.05%) but was better in arm 3 (absolute HbA1c level difference, -0.25%; 95% CI, -0.43% to -0.06%). Total spending and office visits did not differ, but arm 2 (moderate intensity intervention) had more hospitalizations (odds ratio, 1.22; 95% CI, 1.06-1.41) and emergency department visits (odds ratio, 1.38; 95% CI, 1.24-1.53) than did arm 1 (low intensity intervention).

Conclusions and Relevance

Compared with an untargeted low-intensity intervention, delivering a highly targeted high-intensity intervention did not improve insulin persistence but modestly improved mean glycemic control. A partially targeted moderate-intensity intervention did not change insulin persistence or HbA1c level but was associated with a small increase in hospitalizations.

Trial Registration

ClinicalTrials.gov Identifier: NCT02846779.

Provider Perspectives of Medication Complexity in Home Health Care: A Qualitative Secondary Data Analysis

A primary service provided by home care is medication management. Issues with medication management at home place older adults at high risk for hospital admission, readmission, and adverse events. This study sought to understand medication management challenges from the home care provider perspective. A qualitative secondary data analysis approach was used to analyze program evaluation interview data from an interprofessional educational intervention study designed to decrease medication complexity in older urban adults receiving home care. Directed and summative content analysis approaches were used to analyze data from 90 clinician and student participants. Medication safety issues along with provider–provider communication problems were central themes with medication complexity. Fragmented care coordination contributed to medication management complexity. Patient-, provider-, and system-level factors influencing medication complexity and management were identified as contributing to both communication and coordination challenges.

Estimating Adherence Based on Prescription or Dispensation Information: Impact on Thresholds and Outcomes. A Real-World Study With Atrial Fibrillation Patients Treated With Oral Anticoagulants in Spain

Objective: To estimate drug exposure, Proportion of Days Covered (PDC) and percentage of patients with PDC ≥ 80% from a cohort of atrial fibrillation patients initiating oral anticoagulant (OAC) treatment. We employed three different approaches to estimate PDC, using either data from prescription and dispensing (PD cohort) or two common designs based on dispensing information only, requiring at least one (D1) or at least two (D2) refills for inclusion in the cohorts. Finally, we assessed the impact of adherence on health outcomes according to each method. Methods: Population-based retrospective cohort of all patients with Non Valvular Atrial Fibrillation (NVAF), who were newly prescribed acenocoumarol, apixaban, dabigatran or rivaroxaban from November 2011 to December 2015 in the region of Valencia (Spain). Patients were followed for 12 months to assess adherence using three different approaches (PD, D1 and D2 cohorts). To analyze the relationship between adherence (PDC ≥ 80) defined according to each method of calculation and health outcomes (death for any cause, stroke or bleeding) Cox regression models were used. For the identification of clinical events patients were followed from the end of the adherence assessment period to the end of the available follow-up period. Results: PD cohort included all patients with an OAC prescription (n = 38,802), D1 cohort excluded fully non-adherent patients (n = 265) and D2 cohort also excluded patients without two refills separated by 180 days (n = 2,614). PDC ≥ 80% ranged from 94% in the PD cohort to 75% in the D1 cohort. Drug exposure among adherent (PDC ≥ 80%) and non-adherent (PDC < 80%) patients was different between cohorts. In adjusted analysis, high adherence was associated with a reduced risk of death [Hazard Ratio (HR): from 0.82 to 0.86] and (except in the PD cohort) the risk for ischemic stroke (HR: from 0.61 to 0.64) without increasing the risk of bleeding. Conclusion: Common approaches to assess adherence using measures based on days' supply exclude groups of non-adherent patients and, also, misattribute periods of doctors' discontinuation to patient non-adherence, misestimating adherence overall. Physician-initiated discontinuation is a major contributor to reduced OAC exposure. When using the PDC80 threshold, very different groups of patients may be classified as adherent or non-adherent depending on the method used for the calculation of days' supply measures. High adherence and high exposure to OAC treatment in NVAF patients is associated with better health outcomes.

Impact of Medication Adherence on Health Services Utilization in Medicaid

OBJECTIVE

To examine the impact of adherence to chronic disease medications on health services utilization among Medicaid enrollees.

SUBJECTS

Eligibility, claims, and encounter data from the Medicaid Analytic Extract files from 10 states (Alabama, California, Florida, Illinois, Indiana, Louisiana, New Hampshire, New Mexico, New York, and Virginia) were used to construct a 3-year (2008-2010), longitudinal dataset of Medicaid recipients 18-64 years of age, including 656,646 blind/disabled individuals and 704,368 other adults. Patients were classified as having ≥1 of 7 chronic conditions: (1) congestive heart failure; (2) hypertension; (3) dyslipidemia; (4) diabetes; (5) asthma/chronic obstructive pulmonary disease; (6) depression; and (7) schizophrenia/bipolar.

METHODS

Poisson regression was used to estimate associations between medication adherence [continuous and categorical proportion of days covered (PDC)] and 3 dependent variables: number of inpatient hospitalizations, emergency department visits, and outpatient physician/clinic visits.

RESULTS

Full adherence was associated with 8%-26% fewer hospitalizations and 3%-12% fewer emergency department visits among those with congestive heart failure, hypertension, diabetes, and schizophrenia/bipolar. In all analyses, full adherence was associated with up to 15% fewer outpatient physician/clinic visits. Moreover, low and moderate levels of adherence were also related to less health care use.

CONCLUSIONS

Substantial reductions in health services utilization and costs may be realized with improved medication adherence in Medicaid. These benefits begin to accrue at adherence levels below the common 0.80 PDC threshold. Therefore, interventions should focus not just on perfecting moderate adherers, but also on encouraging Medicaid patients with chronic conditions to initiate pharmacotherapy.

A Systematic Review of Medication Adherence Thresholds Dependent of Clinical Outcomes

Background: In pharmacotherapy, the achievement of a target clinical outcome requires a certain level of medication intake or adherence. Based on Haynes's early empirical definition of sufficient adherence to antihypertensive medications as taking ≥80% of medication, many researchers used this threshold to distinguish adherent from non-adherent patients. However, we propose that different diseases, medications and patient's characteristics influence the cut-off point of the adherence rate above which the clinical outcome is satisfactory (thereafter medication adherence threshold). Moreover, the assessment of adherence and clinical outcomes may differ greatly and should be taken into consideration. To our knowledge, very few studies have defined adherence rates linked to clinical outcomes. We aimed at investigating medication adherence thresholds in relation to clinical outcomes.

Method: We searched for studies that determined the relationship between adherence rates and clinical outcomes in the databases PubMed, Embase^Ⓡ and Web of Science™ until December 2017, limited to English-language. Our outcome measure was any threshold value of adherence. The inclusion criteria of the retrieved studies were (1) any measurement of medication adherence, (2) any assessment of clinical outcomes, and (3) any method to define medication adherence thresholds in relation to clinical outcomes. We excluded articles considered as a tutorial. Two authors (PB and IA) independently screened titles and abstracts for relevance, reviewed full-texts, and extracted items. The results of the included studies are presented qualitatively.

Result: We analyzed 6 articles that assessed clinical outcomes linked to adherence rates in 7 chronic disease states. Medication adherence was measured with Medication Possession Ratio (MPR, n = 3), Proportion of Days Covered (PDC, n = 1), both (n = 1), or Medication Event Monitoring System (MEMS). Clinical outcomes were event free episodes, hospitalization, cortisone use, reported symptoms and reduction of lipid levels. To find the relationship between the targeted clinical outcome and adherence rates, three studies applied logistic regression and three used survival analysis. Five studies defined adherence thresholds between 46 and 92%. One study confirmed the 80% threshold as valid to distinguish adherent from non-adherent patients.

Conclusion: The analyzed studies were highly heterogeneous, predominantly concerning methods of calculating adherence. We could not compare studies quantitatively, mostly because adherence rates could not be standardized. Therefore, we cannot reject or confirm the validity of the historical 80% threshold. Nevertheless, the 80% threshold was clearly questioned as a general standard.

Transforming Diabetes Care Through Artificial Intelligence: The Future Is Here

An estimated 425 million people globally have diabetes, accounting for 12% of the world's health expenditures, and yet 1 in 2 persons remain undiagnosed and untreated. Applications of artificial intelligence (AI) and cognitive computing offer promise in diabetes care. The purpose of this article is to better understand what AI advances may be relevant today to persons with diabetes (PWDs), their clinicians, family, and caregivers. The authors conducted a predefined, online PubMed search of publicly available sources of information from 2009 onward using the search terms "diabetes" and "artificial intelligence." The study included clinically-relevant, high-impact articles, and excluded articles whose purpose was technical in nature. A total of 450 published diabetes and AI articles met the inclusion criteria. The studies represent a diverse and complex set of innovative approaches that aim to transform diabetes care in 4 main areas: automated retinal screening, clinical decision support, predictive population risk stratification, and patient self-management tools. Many of these new AI-powered retinal imaging systems, predictive modeling programs, glucose sensors, insulin pumps, smartphone applications, and other decision-support aids are on the market today with more on the way. AI applications have the potential to transform diabetes care and help millions of PWDs to achieve better blood glucose control, reduce hypoglycemic episodes, and reduce diabetes comorbidities and complications. AI applications offer greater accuracy, efficiency, ease of use, and satisfaction for PWDs, their clinicians, family, and caregivers.

Artificial Intelligence for Diabetes Management and Decision Support: Literature Review

Background

Artificial intelligence methods in combination with the latest technologies, including medical devices, mobile computing, and sensor technologies, have the potential to enable the creation and delivery of better management services to deal with chronic diseases. One of the most lethal and prevalent chronic diseases is diabetes mellitus, which is characterized by dysfunction of glucose homeostasis.

Objective

The objective of this paper is to review recent efforts to use artificial intelligence techniques to assist in the management of diabetes, along with the associated challenges.

Methods

A review of the literature was conducted using PubMed and related bibliographic resources. Analyses of the literature from 2010 to 2018 yielded 1849 pertinent articles, of which we selected 141 for detailed review.

Results

We propose a functional taxonomy for diabetes management and artificial intelligence. Additionally, a detailed analysis of each subject category was performed using related key outcomes. This approach revealed that the experiments and studies reviewed yielded encouraging results.

Conclusions

We obtained evidence of an acceleration of research activity aimed at developing artificial intelligence-powered tools for prediction and prevention of complications associated with diabetes. Our results indicate that artificial intelligence methods are being progressively established as suitable for use in clinical daily practice, as well as for the self-management of diabetes. Consequently, these methods provide powerful tools for improving patients’ quality of life.

Adherence to Antipsychotic Medication and Criminal Recidivism in a Canadian Provincial Offender Population

Preliminary evidence suggests that adherence to antipsychotic medication reduces criminal recidivism among patients diagnosed with schizophrenia. However, existing studies operationalize antipsychotic adherence as a binary variable (usually using a threshold of ≥80%), which does not reflect the prevalence of suboptimal adherence in real-world settings. The purpose of the current analysis was to investigate the association between successive ordinal levels of antipsychotic adherence and criminal recidivism in a well-defined sample of offenders diagnosed with schizophrenia (n = 11462). Adherence was measured using the medication possession ratio (MPR) and analyzed as a time-dependent covariate in multivariable regression models. Data were drawn from linked, comprehensive diagnostic, pharmacy and justice system records, and individuals were followed for an average of 10 years. Adjusted rate ratios (ARR) and confidence intervals (CI) are reported. Overall mean MPR was 0.41. Increasing levels of antipsychotic adherence were not associated with progressively lower rates of offending. However, when compared to the reference group (MPR ≥ 80%) all lower adherence levels were significantly associated (P < .001) with increased risk of violent (ARR = 1.58; 95% CI = 1.46-1.71) and nonviolent (ARR = 1.41; 95% CI = 1.33-1.50) offenses. Significance was replicated in separate sensitivity analyses. Previously published studies reporting reductions in crime may have been influenced by antipsychotic adherence ≥80%. Binary operationalization of adherence is an inaccurate predictor of recidivism. Future research addressing functional outcomes of antipsychotic adherence should conceptualize adherence as an incremental independent variable.

Prediction of Emergency Department Hospital Admission Based on Natural Language Processing and Neural Networks

OBJECTIVE

To describe and compare logistic regression and neural network modeling strategies to predict hospital admission or transfer following initial presentation to Emergency Department (ED) triage with and without the addition of natural language processing elements.

METHODS

Using data from the National Hospital Ambulatory Medical Care Survey (NHAMCS), a cross-sectional probability sample of United States EDs from 2012 and 2013 survey years, we developed several predictive models with the outcome being admission to the hospital or transfer vs. discharge home. We included patient characteristics immediately available after the patient has presented to the ED and undergone a triage process. We used this information to construct logistic regression (LR) and multilayer neural network models (MLNN) which included natural language processing (NLP) and principal component analysis from the patient's reason for visit. Ten-fold cross validation was used to test the predictive capacity of each model and receiver operating curves (AUC) were then calculated for each model.

RESULTS

Of the 47,200 ED visits from 642 hospitals, 6,335 (13.42%) resulted in hospital admission (or transfer). A total of 48 principal components were extracted by NLP from the reason for visit fields, which explained 75% of the overall variance for hospitalization. In the model including only structured variables, the AUC was 0.824 (95% CI 0.818-0.830) for logistic regression and 0.823 (95% CI 0.817-0.829) for MLNN. Models including only free-text information generated AUC of 0.742 (95% CI 0.731- 0.753) for logistic regression and 0.753 (95% CI 0.742-0.764) for MLNN. When both structured variables and free text variables were included, the AUC reached 0.846 (95% CI 0.839-0.853) for logistic regression and 0.844 (95% CI 0.836-0.852) for MLNN.

CONCLUSIONS

The predictive accuracy of hospital admission or transfer for patients who presented to ED triage overall was good, and was improved with the inclusion of free text data from a patient's reason for visit regardless of modeling approach. Natural language processing and neural networks that incorporate patient-reported outcome free text may increase predictive accuracy for hospital admission.

Are Randomized Controlled Trials the (G)old Standard? From Clinical Intelligence to Prescriptive Analytics

Despite the accelerating pace of scientific discovery, the current clinical research enterprise does not sufficiently address pressing clinical questions. Given the constraints on clinical trials, for a majority of clinical questions, the only relevant data available to aid in decision making are based on observation and experience. Our purpose here is 3-fold. First, we describe the classic context of medical research guided by Poppers' scientific epistemology of "falsificationism." Second, we discuss challenges and shortcomings of randomized controlled trials and present the potential of observational studies based on big data. Third, we cover several obstacles related to the use of observational (retrospective) data in clinical studies. We conclude that randomized controlled trials are not at risk for extinction, but innovations in statistics, machine learning, and big data analytics may generate a completely new ecosystem for exploration and validation.

Challenges associated with insulin therapy progression among patients with type 2 diabetes: Latin American MOSAIc study baseline data

BACKGROUND

Poor glycemic control in patients with type 2 diabetes is commonly recorded worldwide; Latin America (LA) is not an exception. Barriers to intensifying insulin therapy and which barriers are most likely to negatively impact outcomes are not completely known. The objective was to identify barriers to insulin progression in individuals with type 2 diabetes mellitus (T2DM) in LA countries (Mexico, Brazil, and Argentina).

METHODS

MOSAIc is a multinational, non-interventional, prospective, observational study aiming to identify the patient-, physician-, and healthcare-based factors affecting insulin intensification. Eligible patients were ≥18 years, had T2DM, and were treated with insulin for ≥3 months with/without oral antidiabetic drugs (OADs). Demographic, clinical, and psychosocial data were collected at baseline and regular intervals during the 24-month follow-up period. This paper however, focuses on baseline data analysis. The association between glycated hemoglobin (HbA1c) and selected covariates was assessed.

RESULTS

A trend toward a higher level of HbA1c was observed in the LA versus non-LA population (8.40 ± 2.79 versus 8.18 ± 2.28; p ≤ 0.069). Significant differences were observed in clinical parameters, treatment patterns, and patient-reported outcomes in LA compared with the rest of the cohorts and between Mexico, Brazil, and Argentina. Higher number of insulin injections and lower number of OADs were used, whereas a lower level of knowledge and a higher level of diabetes-related distress were reported in LA. Covariates associated with HbA1c levels included age (-0.0129; p < 0.0001), number of OADs (0.0835; p = 0.0264), higher education level (-0.2261; p = 0.0101), healthy diet (-0.0555; p = 0.0083), self-monitoring blood glucose (-0.0512; p = 0.0033), hurried communication style in the process of care (0.1295; p = 0.0208), number of insulin injections (0.1616; p = 0.0088), adherence (-0.1939; p ≤ 0.0104), and not filling insulin prescription due to associated cost (0.2651; p = 0.0198).

CONCLUSION

MOSAIc baseline data showed that insulin intensification in LA is not optimal and identified several conditions that significantly affect attaining appropriate HbA1c values. Tailored public health strategies, including education, should be developed to overcome such barriers. Trial Registration NCT01400971.