About Multiple Readmissions Predictive Model

The Multiple Readmissions Predictive Model uses a combination of clinical theory and machine learning to automatically identify patients who are likely to be readmitted multiple times over the next year.

The model identifies high-risk patients for enrolment into the Hospital to Home programme. Through the programme, nurses visit patients and their caregivers at their homes, educate them on caring for themselves, and help them with their needs such as arranging for meals or changing of wound dressings.

Singapore’s public hospitals admit more than 450,000 patients every year. The use of the model reduces the average daily vetting workload by 10 to 15 per cent of the national average of daily admissions of about 1,200 patients per day.

Improves care support

Potentiall reduces the length of patient’s hospital stays

Increases productivity for staff


Multiple Readmissions Predictive Model at a glance


Traditionally, care support programmes identify patients who are at risk of multiple hospital readmissions through historical data. These risk scoring methods have limitations in handling patients with multiple co-morbidities. 

Nurses previously had to spend almost half a day manually screening through the inpatient list and speaking to newly-admitted patients and their care team to identify those at risk of multiple readmissions.


The Multiple Readmissions Predictive Model can analyse multi-dimensional facets of the patient, ranging from chronic diseases to end-of-life conditions, and has an prediction accuracy of seven in 10 patients. Using over a thousand indicators, including patient age, number of inpatient admissions and total length of stay in the past two years, the model automatically identifies patients who have a history or are at risk of multiple readmissions.


The early identification of high-risk patients allows for timelier intervention to help reduce patients’ average length of stay in the hospitals and healthcare utilisation. Nurses, doctors and clinicians can focus on the smaller group of high-risk patients flagged out by the model for the second round of vetting. As the care team no longer needs to screen the remaining 85-90 per cent of patients, they can focus more time on direct patient care.

Future Outlook

The next step is to move the Multiple Readmissions Predictive Model further upstream to primary care settings, to enable early intervention and delay the progression of patients’ conditions.


Implementation of Predictive Analytics

  • Business integration approach
    Developed as part of the Hospital to Home (H2H) service involving clinicians, care team, policy maker and administrator.
  • Creation of analytics infrastructure
    Business Research Analytics Insights Network (BRAIN) deployed as national/common analytics platform to support the Ministry, agencies, Public Health Institutions (PHIs) by:
    i. Bringing data together
    ii. Linking patients 
    iii. Harmonising data
    iv. Anonymising data
    v. Providing common toolsets and capabilities
    vi. Collaborating in secured environment
  • Multi-disciplinary skill sets
    • Respiratory and critical care
    • Family medicine and continuing care
    • Bio-informatics
    • Health economics
    • Health delivery
    • Health research
    • Data science and analytics
  • Agile and scientific methodology

    Scientific and systematic development approach to gain the trust of user.

  • Volume and veracity of data
    • 1.4 billion data points
    • 7 million records
    • 200+ variables
    • Automated generation of prediction risk score using three-year data on a daily basis



By continuing to use and navigate this website, you consent to the use of cookies in accordance with our Privacy Policy.