Several years ago, when I was a climate data scientist, I received an email from my brother, who is a neurosurgeon, with my mom’s chest X-ray images as attachments. He asked me to find a doctor in the US to help settle the discrepancy of several doctors’ different diagnosis of my mom’s mysterious coughing. For the following couple of months, I still remember the distress of wishing to be able to read those images like a radiologist and tell my mom and family that there was nothing to worry about.
But she was later diagnosed with Pulmonary Fibrosis, and her chest images were sent to the doctor every six months for check ups. I later made my career change from a climate data science to a more general data scientist, and healthcare predictive analytics has become one of my passions.
I used to work for an organization combining healthcare, education, and pension for home care givers. There, I managed several predictive intervention projects partnering with multiple cross-functional teams. One of the projects, no-show predictive intervention, reduced 40% of care giver churn within six months. We contribute the effectiveness of this project to the fact that it bridged the gaps between data insights and measurable interventions. It pipelined the multiple data sources governance, predictive modelling, and intervention evaluations. A set of best practices to intervene at-risk no-shows can be recommended and scaled for new care givers.
For clinical no-shows, on average, it is as high as a 27% rate across healthcare systems in North America. High no-show rates have been identified as one of the most significant barriers to access care for the underserved and uninsured populations. This population typically has significant chronic mental and health conditions that become especially costly when routine treatment is not maintained, therefore they tend to no-show. High no-show rates pose a great challenge for the healthcare system.
Many intervention tactics have been experimented to reduce no-shows, such as letters, emails, texts, phone calls, as well as overbooking strategies. These solutions come with their own drawbacks and extra costs. Moreover, the impact of these intervention options needs to be evaluated and customized for specific clinics and patients.
Our preliminary research shows that with 1% of no-show rates improvement, a clinic with an average of 15% no-show rates is expected to save $7,000 per month. In other words, about $70,000 will be saved per month for 10 similar clinics, or the same amount of money will be saved ten months for one single clinic.
At Blueprint, our team is passionate about applying our expertise in machine deep learning and in healthcare in order to tackle top challenges that many healthcare organizations are facing (such as clinical no-shows, provider burnouts, pharmacy/clinics compliance, and digital diagnosis). We thrive to provide end-to-end services and solutions to fix the pain points throughout the predictive intervention process.
