Clinical decision support systems (CDSSs) are useful for providing information and knowledge to improve diagnosis and treatment outcomes and to elevate healthcare quality for patients at clinical settings. Thus, CDSSs have been widely applied in medical event detection, disease treatment and management, and drug-dosing or medication-prescribing. In previous studies, we designed clinical predictive models to predict events, including re-admissions, erectile dysfunctions (EDs), and acute myocardial infarction (AMI) based on the data extracted from the National Health Insurance Research Database (NHIRD). A variety of associated comorbidity diseases and comorbidity-related features such as follow-up durations and annual visits to physicians of individual commodities, were included for predictive models to detect high-risk patient for early intervention. Diseases may be caused or deteriorated by other comorbid diseases. In this case, the diseases were not well managed. For example, the annual visits to physicians of most related comorbid diseases for ED patients were significantly lower than those of non-ED patients. Similar results were also observed by comparing AMI and non-AMI patients. These observations implied that improper disease management with less frequent visits to physicians was associated with comorbidity deterioration, resulting in the developments of ED or AMI. Therefore, proper comorbidity management and monitoring is important in preventing comorbidity deterioration and avoiding development of other diseases. In recent years, the leading causes of morbidity and mortality have changed from infectious diseases and malnutrition to non-communicable chronic diseases. Poor lifestyles, including inappropriate dietary pattern, sedentary activity, smoking, alcohol or narcotic indulgence, and chronic stress, are the main factors causing the chronic diseases. For example, according to clinical experiences, modification of behavior or lifestyle was shown to be effective in improving symptoms for patients with chronic disease, such as interstitial cystitis/bladder pain syndrome (IC/BPS) patients. In our previous reports, changing the daily behaviors by weekly health education through a text-based e-Health and video-based m-Health systems was shown to be effective in alleviating pains and other uncomfortable symptoms and improving quality of life for women with IC/BPS. Recently, Kao et al. also proposed an m-Health system for patients with primary hypertension to self-titrate their blood pressure with significant improvement in systolic and diastolic blood pressure only after 3 months of intervention. According to the aforementioned description, this study was carried out to apply a high-performance predictive model to detect high-risk patients for early interventions as well as to design an m-Health system for providing medical educations and reminding regular visits to physicians to manage their comorbidities. The purpose of the study was to prevent the deterioration of associated diseases, which in turn reduced the ED occurrences.

A. Predictive Models A graphic user’s interface (GUI) was designed to be used in the urological clinic based on the best model by comparing the models developed using IGS and DNN algorithms. Data of the candidate male patients visiting Dr. Wang’s (a urologist) clinics were collected from the electronic medical record of Chi-Mei Hospital for input to the predictive model. The candidate patients were also requested to fill a questionnaire regarding diagnosed ages, follow-up durations, and annual visits to physician for related comorbidities which might not be included in medical records of Chi Mei medical center. The patient data for the predictive model included patient’s current age, the presence of 10 comorbidities, and 30 comorbidity-related feutures (diagnosed ages, follow-up durations, and annual visits to physicians of individual comorbidities). Patients with a probability higher than 50% determined by the ED predictive model and confirmed by the physician were recruited as participants for this study. B. Design of M-Health System The materials for health educational of various comorbidities were prepared based on the guidelines of healthy lifestyles and behaviors. Existing materials were also available on the internet. All the educational materials were included in the m-Health system after being determined and selected by the urologist, who has served as the attending physician of the participants in Chi Mei Medical Center. Reminding messages were sent to the studied participants from the m-Health system to call them back to the clinics for monitoring progression of the comorbidities within a week before and after the registered date of visiting physicians. Figures 1 and 2 demonstrate the APP interfaces for participants and case manager, respectively. C. Participants and Experimental Protocols A total of 31 patients with higher probability (>50%) in developing ED were recruited from the urological clinic of Chi-Mei Medical Center. After signing the informed consent with a brief introduction about how to use the m-Health APP, the experiment began immediately by assessing baseline ED severity based on the patient’s medical record, recent symptoms, and scores of Sexual Health Inventory for Men (SHIM) questionnaire. After 6 months of m-Health intervention, the participants were asked to fill the SHIM questionnaire again to compare with the the baseline assessment. In addition, participants’ intention to use the m-Health system were also surveyed using technology acceptance model (TAM) questionnaire.

A. SHIM Assessment Tables I and II compare the SHIM scores between baseline and post-intervention after 6 months intervention for the study and control groups, respectively. As indicated in Tables I, the mean age of participants was 59.83±8.40 years old and the SHIM scores at the baseline and after m-Health intervention were 12.70±8.45 and 14.78±8.96, respectively, with significant improvement (pair t-test, p<0.05) after 6 months of ED m-Health intervention. Other the other hand, Tables II shows that the SHIM scores at the baseline and after m-Health intervention were 16.73±6.49 and 16.77±6.42, respectively, without significant improvement (pair t-test, p=0.90) after 6 months of observation. Table III compares the SHIM score between 2 groups after 6 months of intervention. No significant difference (Independent t-test, p>0.05) is found in baseline, post-intervention, or improvement between 2 groups. However, if we take the number of improved and non-improved patients into account, significant difference (Chi-square test, p<0.001) in odd ratio is observed between 2 groups. The ratio of improved to non-improved patients of the study group is significant higher than the control group, indicating m-Health is effective in improving sexual function of the patients.

As shown in Table I, the SHIM score after m-Health intervention was significantly higher than the baseline (14.78±8.96 vs 12.70±8.45; pair t-test, p<0.05), indicating that ED m-Health system was useful in improving ED symptoms. The result was validated by the PU of the TAM survey, which exhibits that most participants agreed the m-Health APP was useful in managing their comorbidities (one sample t-test, p<0.001) and alleviating their ED symptoms (p<0.001). Their QoL (p<0.001) was also improve. The result was consistent with our previous studies with e-Health and m-Health systems adopted for managing patients with IC/BPS. Text-based and video-based health educational materials were adopted for modifying patients’ behavior and life-style as well as for managing emergent symptoms to prevent their symptoms. In contrast, in addition to health education, managements of 10 comorbidities related to ED occurrence were also integrated in the m-Health APP for reminding patients to regularly visit physicians to manage their comorbid diseases. By analyzing annual visits to physicians of comorbidities, we observed that unmanaged comorbidities might induce ED occurrence. Hence, reminding patients to have regular visits to physicians a few days before the appointment date is useful for managing their comorbidities, which in turn prevents disease deterioration A range of 25 to 254 days with a median of 66 days was long enough for developing a good habit (Lailly et al. 2010). In this study, the intervention duration was more than 6 months, which was long enough for consolidating healthy behaviors and lifestyles to prevent deterioration of relevant comorbidities and is effective in improving ED symptom