Overcrowded Hospitals --- and Bottlenecks
ROI=Zero if you improve anything other than a true bottleneck
Two weeks ago, Kate Gamble started
a terrific blog topic on Overcrowding at Hospitals. The posting is worth going back and reading. Kate provides the data and links that really describe the overcrowding situations well. In my blog on
ER wait times, I too elaborated the problem, the CDC data comparing now to ten years ago, and we addressed some solutions. Kate solicited input on "
patient flow systems", as part of the solution.
One smart reader, Julie E. Crowe, chimed in and offered the perspective that, at least in the ED, overcrowding is not a problem that can be solved by a CIO. I was comforted by that response, and it was concordant with my earlier blog on the topic. That said, what can a CIO, or CNO, COO, or CMIO, do to address the problem Kate so eloquently described?
Process Timing - a BI story
Working with Shakespeare Health, a pseudonym for a real, multi-hospital system in the Northeast, I co-led a team that asked exactly those questions: how can an EMR directly enhance patient care, safety, and throughput. To look at patient flow, clinical processes, clinical severity, overcrowding, and economics, we drew data from multiple systems managed by the CIO and his IT department. These included ADT, a clinical system (including lab, pharmacy, nursing, and co-existing illness data), several financial systems, including UB-92 claims and all charges, and data from a commercial, dedicated clinical profiling and outcomes system.
Here's what the conceptual model looked like: (exemplary ED->ICU->Floor path)
The most striking results were the process times, and how they informed the management options. For example, for the lowest severity patients, the median time spent in Resource Center A was 276 minutes. The average was well over twice that (not surprisingly). The median times for the all other severities was consistently in the 4 to 5 hour range.
Sensitivity analyses and other calculations showed that the distributions were not driven by a few 'outliers' but represented true bottlenecks. That's where the CXOs can determine is a patient flow system is the solution to overcrowding, or some other fixed resource constraint is driving the system behavior. To Julie's point, in the later case, a patient flow system wont change throughput.
The complete model, populated graphically with the process data and management actions is here:
This picture summaries the time spent in each care setting (E R, Floor, ICU), for each flow. For example, the top, left-most chart labeled "floor" were direct admission that did not go through the ED. This picture was used as part of an internal presentation to communicate the results.
In contrast with other approaches, this retrospective analysis revealed the annual monetary impact would be on Shakespeare Health for each of the initiatives listed. They are "Re-direct low risk patients ...", increasing the size of the observation unit, instituting new pathways, more aggressively using case managers than historical experience, etc.
These results were calculated based on actual payer mix implications on reimbursements. Scenario's were evaluated factoring in hospital census to calculate the capacity implications (e.g. overcrowding, diversion, and economic impact of not having a medical bed, against it's expected profitability.) The actions were also normalized to factor in the seasonality of overcrowding.
Process Timing leads to Bottleneck Management
The work described above was initially done using Bottleneck Management principle elaborated in an extremely popular book in it's time, The Goal, by Eli Goldratt. The book focuses on the management and social issues. It sneaks in inventory management, operating expenses, cost accounting, scheduling and other domains, in the context of a story.
The big take-aways for Overcrowded Hospitals are very simple:
1) If your process has a bottleneck, and you improve anything about the process other than the bottle, the bottleneck will guarantee that the throughput will be unchanged.
2) Work-in-process inventory will reliably build up in front of the bottleneck. Any good electronic transactional system you already have in place can probably see it today. I included the graphics here to provide an example. You don't need to put in a new patient flow system to see it; how you manage the bottleneck depends on what you see. Julie was right!
3) Generally, you can't eliminate bottlenecks, although that would be nice. You need to exploit them and subordinate other processes. That's beyond this blog post. If asked, I will elaborate; comment below for a public answer, or email for a private response.
4) In healthcare processes, especially in hospitals, during the course of a year, bottlenecks definitely do "float." So, inadequate numbers of observation beds in February was a bottleneck at Shakespeare; it wasn't in June.
5) The local medical staff needs to have a central role in this work. Interpreting clinical process data without a deep grounding in diagnostic and therapeutic context is hazardous to anyone's health. So is proposing changes in care delivery systems without being able to credibly speak to that context.
6) The ROI side of the story: Capacity-driven organizations, that is, high fixed-cost, low variable cost operations are primarily driven by plant utilization. Improving operations, only to create capacity you don't use does not produce ROI. There is real, CFO-certifiable ROI in driving up 'throughput-dollar-day' plant utilization, by improving clinical operations as outlined here. This was the case at Shakespear. This can be a great management opportunity for overcrowded hospitals. In our example, the improvement in pneumonia management allowed for more profitable utilization of medical beds, while maintaining (and in some cases improving) clinical performance quality.
What experience do we have in the readership regarding patient throughput and related BI initiatives?