Using Machine Learning to Optimize Hospital Operations
Many healthcare systems grapple with the paradox where assets are over-booked but under-utilized daily. LeanTaaS is a startup that is using machine learning to solve this problem.
There has been a lot of excitement around the application of machine learning (ML) to transform the healthcare sector. An Accenture report predicts that the AI health market will grow at CAGR of 40% to reach $6.6B by 2021. However, machine learning in healthcare is not new – for example, MYCIN was a rules-based machine learning system developed at Stanford in 1978 that was shown to diagnose bacterial infections better than pathologists; nevertheless, it did not change medical practices. Today, only 5% of health systems leverage AI/ML. But we may be at a tipping point. As healthcare transitions from a fee-for-service to fee-for-value reimbursement system and the amount of digital data (EHR, imaging, genomics, wearables) proliferates, healthcare systems now have a strong incentive to invest in robust data infrastructure, acquire data management expertise, and embrace AI/ML solutions.
One of the areas where ML may realize its biggest impact in healthcare in the short-term is automating mundane tasks for physicians and hospital administrators. LeanTaaS is a software company that uses lean principles and machine learning to help hospitals better allocate hospital resources and staff efficiently.
One of LeanTaaS’ first use cases is in operating rooms (ORs). Traditionally, most hospitals schedule ORs using block scheduling, where surgeons are assigned a block of time for their procedures. While this type of scheduling is appropriate for businesses such as fitness studios where the start and end of a session are defined, healthcare procedures have a lot of variability in the start and end of its sessions (e.g., a surgeon only requires half the allotted time to finish a procedure) which necessitates a more flexible system. Exacerbating this problem is the fact that rescheduling is currently done via a cumbersome process, where a surgeon calls/faxes/emails the administrators and barter for open blocks. Coordinating the staff, supplies, and medical equipment for the procedure adds further friction to rescheduling. By various reports, OR delays occur in 40% to 96% of cases and are very costly. The variability of medical procedures causes some ORs to be idle or blocked at various times during the day, contributing to the asset utilization paradox – where assets are over-booked but under-utilized on a daily basis.
Figure 1 : OR block scheduling in Grey’s Anatomy
LeanTaaS’ iQueue platform addresses this paradox by predicting the demand patterns of hospital ORs using machine learning. LeanTaaS’ algorithm uses historical EHR data and staff schedules to aggregate treatment types and generate block templates taking into account the forecasted volume and mix of procedures, and surgical block requirements per service line and individual surgeons. The product becomes more robust over time as machine learning is used to continuously improve the accuracy of the forecasts and the quality of the recommended block allocation. LeanTaaS embeds these predictive analytic capabilities in a user-friendly scheduling app that allows surgeons and their schedulers to easily request and release OR blocks.
Figure 2: LeanTaaS iQueue Platform Schematic
Beyond ORs, LeanTaaS has also developed products for infusion centers and labs to reduce patient wait times and eliminate bottlenecks. With $39M raised from Insight Ventures and Sedgwick Claims Management Services, LeanTaaS is now used by over 40 providers across the country.
To become the leading healthcare operations platform, LeanTaaS should consider a few potential next steps. In the short-term, LeanTaaS should place more weight on developing and cross-selling a full suite of products to existing hospital systems rather than selling to new customer because 1) hospital sale cycle is notoriously slow, and 2) the value of LeanTaaS’ platform increases as hospitals deploy the software to more systems because the utilization of staff and resources are linked across these services and hospital administrators will be able to better make decisions holistically rather than in silos. In the medium term, LeanTaaS will be able to increase the value of the cost and utilization data its platform currently captures by also integrating quality data. Quality data, such as readmission rate after a procedure, when combined with cost and utilization data will enable hospitals and physicians to make even better operational and clinical decisions. In the long term, LeanTaaS could consider building additional products that capture the patient experience or enable hospitals to rent resources from different systems.
As LeanTaaS scales, how should it build a defensible business against tech behemoths such as IBM Watson Health that have amassed massive amount of data?
 Matt Collier et al., “Artificial Intelligence: Healthcare’s New Nervous System”, Accenture (2017), https://www.accenture.com/us-en/insight-artificial-intelligence-healthcare
 Vijay Pande et al., “Putting AI in Medicine, in Practice”, a16z podcast, https://a16z.com/2017/11/03/ai-medicine/
 Megan Zweig et al., “Demystifying AI and Machine Learning in Healthcare”, Rock Health report
 Jacqueline LaPointe, “OR Efficiency, Machine Learning Boosts UCHealth’s Revenue by $10M”, https://revcycleintelligence.com/news/or-efficiency-machine-learning-boosts-uchealths-revenue-by-10m
 Sanjeev Agarwal, “Why Hospitals Need Better Data Science”, HBR, https://hbr.org/2017/10/why-hospitals-need-better-data-science
Student comments on Using Machine Learning to Optimize Hospital Operations
This is a fascinating use of machine learning in the healthcare space. There is certainly a need for a better solution than what most hospitals are currently using, and this may very well be it. However, as with most things in healthcare, there is a lot of complexity beneath the surface. I would imagine that data related to several other inputs (e.g. availability, individual variability in efficiency, etc. of cleaning staff required between OR cases) are either currently not collected or are collected in an entirely different system (e.g. a timekeeping system) than the OR case scheduling system. With these factors (e.g. is the person assigned to clean the room generally fast or slow at their job) playing a significant role in turnaround time between cases and general efficiency of the OR’s, there is a lot of work required at each site to turn LeanTaaS into a comprehensive solution. If they can figure out how to do that at a reasonable cost, I imagine that they will quickly build a huge book of business.
Really interesting write-up! In this case, since a lot of the data used to generate the schedules are specific to the hospital, perhaps that levels the playing field a bit and reduces the data advantage that the behemoths have against LeanTaaS? Additionally, I think the user-friendly application you mentioned could help add some defensibility and increase switching costs. I totally agree on the cross-selling as well, which also could have another benefit of embedding LeanTaaS further in a hospital and reducing churn.