Primary Care Measurement:  Lake or Stream?

Primary Care Measurement: Lake or Stream?


Last year, I worked on a project to help organizations become ‘Conversation Ready’—to help organizations engage, steward and respect information about end-of-life wishes and intentions.  The project was designed to complement work in communities to help individuals and families have ‘The Conversation’ (here’s the link to The Conversation Project).

One of the project measurements shared by the organizations was ‘Percent of people with documentation of “What Matters Most”.’  The denominator is the count of people served by the organization; the numerator is the count of people in the denominator with documentation of “What Matters Most."

This is an example of a ‘process’ measurement, a measure of how care is provided.  I distinguished between two types of process measure in this post.  The rest of this post relates especially to adherence-type process measures.

My IHI colleague Lauren Macy and I looked at process measurements from a range of organizations and realized we had two types of organizations.  We named those two types ‘Lake’ and ‘Stream.’

In our project, we asked teams to report progress each month, so I’ll discuss our thinking in terms of monthly measurement cycles.

Lake Organizations Have Fixed Denominators, in the short term

A fixed denominator is typical of a residential facility.  The organization provides frequent and repeated services to people in the denominator month after month.  When the organization counts the number of people with documentation of “What Matters Most,” they will base their counts on roughly the same set of people, every month.   It’s ‘roughly the same’ because there will be a small fraction of people who enter or leave the care of the organization in a month.  Of course, over a few years, the denominator will represent a different set of people, so the organization might be described as a lake that has water coming in and water draining away.

Stream Organizations Have Fluid Denominators, in the short term

A fluid denominator is typical of an acute care service, like an ICU or a medical unit in a hospital.  When the organization counts the number of people with documentation of “What Matters Most,” they count the people treated in the measurement month, roughly a new set of people, month to month.  It’s ‘roughly a new set’ because there may be a small fraction of people who are treated in a calendar month who are still in care in the following month.  For measurement of a fluid population, it is simplest to count people who enter care during the month and determine how well the organization engages, documents, and acts on “What Matters Most.”

Primary Care:  Lake or Stream?

Primary care practices look like a lake organization.  The practice can identify people it considers patients, for example patients linked to care team panels.  The practice provides care to these patients.  Month to month, a typical panel is roughly the same, although a small number of patients may be added and a small number of patients may leave.

To measure the percent of patients receiving specific care interventions, like engagement in “What Matters Most", the denominator is the set of people considered patients.

If engagement with patients in “What Matters Most” is new for the practice, the percent of people with evidence of this care will be very low when the practice first changes care and workflow.  The percent will grow over time. The pace of growth will be fast or slow, depending on how the practice reaches out to patients and delivers the new standard care.

What if we think about primary care as a stream organization?  Then we would measure performance on the patients seen by the practice each month and not dilute the arithmetic by using the entire lake as the denominator. 

If the aim of measurement is to monitor process performance to understand impact of new changes and drive improvement, here’s my argument to use stream arithmetic for primary care adherence measures:

1.      You can find out faster whether your changes represent reliable practice, e.g. reaching 95% or more of patients with the standard of care.

2.      You can track whether the practice provided the new standard of care for each patient, tallying counts on a whiteboard or spreadsheet, day by day. Daily tracking and feedback drive rapid process improvement, treating each day as a Plan-Do-Study-Act cycle.  Just add the daily numbers at the end of the month and you have the stream measurement.

3.      You can improve care with the patients who show up; in the beginning of your work, you don’t need to invest extra effort to reach out to people to get the new care.

4.      The people seen by the practice during a given month tend to be more in need of care than the average person in the primary care lake.  With stream measurement, you focus on people with more need for care.

One weakness of the stream measure is it doesn’t encourage finding and treating patients in the lake who don’t engage with care teams in the practice.  You don’t see the global opportunity in the monthly stream measure:   You could have more than 95% of patients in the stream each month with the new service but if 20% of the patients are never seen, you may be fooled into thinking you have made more progress than you really have.  


If your primary care organization has lots of measurement and improvement capacity, you can run the numbers for adherence measures using both the stream and lake arithmetic.  Most organizations don’t have spare capacity, so I recommend that you start by treating primary care as a stream organization for adherence measures.

Do clinical changes cause better outcomes for patients with diabetes?

Do clinical changes cause better outcomes for patients with diabetes?

Self-control and The Model for Improvement

Self-control and The Model for Improvement