Search

BackgroundSMS text messaging programs are increasingly being used for secondary prevention, and have been shown to be effective in a number of health conditions including cardiovascular disease. SMS text messaging programs have the potential to increase the reach of an intervention, at a reduced cost, to larger numbers of people who may not access traditional programs. However, patients regularly reply to the SMS text messages, leading to additional staffing requirements to monitor and moderate the patients’ SMS text messaging replies. This additional staff requirement directly impacts the cost-effectiveness and scalability of SMS text messaging interventions. ObjectiveThis study aimed to test the feasibility and accuracy of developing a machine learning (ML) program to triage SMS text messaging replies (ie, identify which SMS text messaging replies require a health professional review). MethodsSMS text messaging replies received from 2 clinical trials were manually coded (1) into “Is staff review required?” (binary response of yes/no); and then (2) into 12 general categories. Five ML models (Naïve Bayes, OneVsRest, Random Forest Decision Trees, Gradient Boosted Trees, and Multilayer Perceptron) and an ensemble model were tested. For each model run, data were randomly allocated into training set (2183/3118, 70.01%) and test set (935/3118, 29.98%). Accuracy for the yes/no classification was calculated using area under the receiver operating characteristics curve (AUC), false positives, and false negatives. Accuracy for classification into 12 categories was compared using multiclass classification evaluators. ResultsA manual review of 3118 SMS text messaging replies showed that 22.00% (686/3118) required staff review. For determining need for staff review, the Multilayer Perceptron model had highest accuracy (AUC 0.86; 4.85% false negatives; and 4.63% false positives); with addition of heuristics (specified keywords) fewer false negatives were identified (3.19%), with small increase in false positives (7.66%) and AUC 0.79. Application of this model would result in 26.7% of SMS text messaging replies requiring review (true + false positives). The ensemble model produced the lowest false negatives (1.43%) at the expense of higher false positives (16.19%). OneVsRest was the most accurate (72.3%) for the 12-category classification. ConclusionsThe ML program has high sensitivity for identifying the SMS text messaging replies requiring staff input; however, future research is required to validate the models against larger data sets. Incorporation of an ML program to review SMS text messaging replies could significantly reduce staff workload, as staff would not have to review all incoming SMS text messages. This could lead to substantial improvements in cost-effectiveness, scalability, and capacity of SMS text messaging–based interventions.

BackgroundThere are a growing number of mobile phone apps available to support people in taking their medications and to improve medication adherence. However, little is known about how these apps differ in terms of features, quality, and effectiveness. ObjectiveWe aimed to systematically review the medication reminder apps available in the Australian iTunes store and Google Play to assess their features and their quality in order to identify high-quality apps. MethodsThis review was conducted in a similar manner to a systematic review by using a stepwise approach that included (1) a search strategy; (2) eligibility assessment; (3) app selection process through an initial screening of all retrieved apps and full app review of the included apps; (4) data extraction using a predefined set of features considered important or desirable in medication reminder apps; (5) analysis by classifying the apps as basic and advanced medication reminder apps and scoring and ranking them; and (6) a quality assessment by using the Mobile App Rating Scale (MARS), a reliable tool to assess mobile health apps. ResultsWe identified 272 medication reminder apps, of which 152 were found only in Google Play, 87 only in iTunes, and 33 in both app stores. Apps found in Google Play had more customer reviews, higher star ratings, and lower cost compared with apps in iTunes. Only 109 apps were available for free and 124 were recently updated in 2015 or 2016. Overall, the median number of features per app was 3.0 (interquartile range 4.0) and only 18 apps had ≥9 of the 17 desirable features. The most common features were flexible scheduling that was present in 56.3% (153/272) of the included apps, medication tracking history in 54.8% (149/272), snooze option in 34.9% (95/272), and visual aids in 32.4% (88/272). We classified 54.8% (149/272) of the included apps as advanced medication reminder apps and 45.2% (123/272) as basic medication reminder apps. The advanced apps had a higher number of features per app compared with the basic apps. Using the MARS instrument, we were able to identify high-quality apps that were rated as being very interesting and entertaining, highly interactive and customizable, intuitive, and easy to use and to navigate as well as having a high level of visual appeal and good-quality information. ConclusionsMany medication reminder apps are available in the app stores; however, the majority of them did not have many of the desirable features and were, therefore, considered low quality. Through a systematic stepwise process, we were able to identify high-quality apps to be tested in a future study that will provide evidence on the use of medication reminder apps to improve medication adherence.

BackgroundMobile health (mHealth) has huge potential to deliver preventative health services. However, there is paucity of literature on theoretical constructs, technical, practical, and regulatory considerations that enable delivery of such services. ObjectivesThe objective of this study was to outline the key considerations in the development of a text message-based mHealth program; thus providing broad recommendations and guidance to future researchers designing similar programs. MethodsWe describe the key considerations in designing the intervention with respect to functionality, technical infrastructure, data management, software components, regulatory requirements, and operationalization. We also illustrate some of the potential issues and decision points utilizing our experience of developing text message (short message service, SMS) management systems to support 2 large randomized controlled trials: TEXT messages to improve MEDication adherence & Secondary prevention (TEXTMEDS) and Tobacco, EXercise and dieT MEssages (TEXT ME). ResultsThe steps identified in the development process were: (1) background research and development of the text message bank based on scientific evidence and disease-specific guidelines, (2) pilot testing with target audience and incorporating feedback, (3) software-hardware customization to enable delivery of complex personalized programs using prespecified algorithms, and (4) legal and regulatory considerations. Additional considerations in developing text message management systems include: balancing the use of customized versus preexisting software systems, the level of automation versus need for human inputs, monitoring, ensuring data security, interface flexibility, and the ability for upscaling. ConclusionsA merging of expertise in clinical and behavioral sciences, health and research data management systems, software engineering, and mobile phone regulatory requirements is essential to develop a platform to deliver and manage support programs to hundreds of participants simultaneously as in TEXT ME and TEXTMEDS trials. This research provides broad principles that may assist other researchers in developing mHealth programs.

Diabetic retinopathy (DR) is the most common microvascular complication of diabetes and is strongly linked with systemic vascular comorbidities. This study investigated if DR predicts risk of cardiovascular disease (CVD) mortality in a high CVD risk cohort. This was a prospective cohort study of 1582 adult participants who presented to a tertiary Australian hospital for evaluation of acute coronary syndrome by coronary angiography. Participants were concurrently examined for DR from mydriatic fundus photographs which were mask-graded according to International Clinical Classification categories of no DR, mild non-proliferative DR, moderate-to-severe NPDR, and proliferative DR. Coronary artery disease was graded from coronary angiograms using the Gensini score. CVD mortality follow-up was assessed 9 years after baseline examination using linkage with the Australian National Death Index. At baseline, 355 (22.4%) participants had any DR. There were 181 (11.4%) fatal CVD events after 9-years follow-up. After controlling for age, sex, BMI, diabetes, total cholesterol, smoking status, hypertension, previous myocardial infarction and stroke, any DR was associated with 1.8-fold higher risk of CVD mortality (Hazard Ratio [HR] 1.84, 95% Confidence Interval [95% CI: 1.30–2.61]). Mild non-proliferative DR (1.85 [1.26–2.72]) and proliferative DR (5.27 [2.32–12.00]) were associated with greater CVD mortality risk. Further adjustment for coronary artery disease using Gensini scores and excluding patients without diabetes had minimal impact on the association. The increased risk of CVD mortality was significant in both men (2.25 [1.60–3.19]) and women (2.38 [1.24–4.58]) with any DR. In individuals with high CVD risk, presence of DR independently predicts increased CVD mortality. This likely reflects additional contribution of microvascular disease to CVD mortality. Individuals with DR may benefit from a comprehensive cardiovascular risk assessment, lifestyle changes, more intensive cardiovascular management and follow-up to minimise risk of death from CVD events. [ABSTRACT FROM AUTHOR]