Shared Baselines as a Guide to Protocols

Paul Levy 1There have been some interesting and important discussions flying across the web in recent days on the issue of protocols in helping to reduce variation and reduce the incidence of harm to patients.  My mistake in the debate was assuming that medical leaders would be reasonable about how protocols should and should not be used.

A doctor friend, highly committed to patient safety, notes:

My point about the protocols is that I have been chastised for not following them in situations where it was blatantly obvious that they did not apply. (“The protocol is there for a reason.”)

The chastisement comes not from hospital administrators, but from clinician leaders in the doctor’s own department:

We just got another email scolding us for not following the “colorectal pathway” sufficiently. One of the provisions of that pathway, for example, is strict limitation of iv fluids, sometimes difficult to “comply” when patients are severely dehydrated from their bowel preps, particularly the elderly.

The initial goals were to decrease opiate use and decrease PACU LOS, both worthy goals, but we’re all annoyed at being beaten over the head with them and getting our hands slapped if we deviate, even with good reason.

It’s ironic that on the one hand we are extolling the virtues of gene-based individualized therapies, but on the other hand we are trying to pigeon-hole every patient into a standardized protocol.


This is disappointing in so many ways, but especially because the solution is in the hands of the profession.  Brent James discussed the sensible application of protocols to clinical process improvement, as employed in the Intermountain Health system, several years ago:

The concept of “shared baselines” came to rule:

1 — Select a high priority clinical process;

2 — Create evidence-based best practice guidelines;
3 — Build the guidelines into the flow of clinical work;
4 — Use the guidelines as a shared baseline, with doctors free to vary them based on individual patient needs;
5 — Meanwhile, learn from and (over time) eliminate variation arising from the professionals, while retain variation arising from patients.

Note that this approach demands [my emphasis] that doctors modify shared protocols on the basis of patient needs.  The aim is not to step between doctors and their patients.  This is very different from the free form of patient care that exists generally in medicine.  Notes Brent, “We pay for our personal autonomy with the lives of our patients.  This is indefensible.”  The approach used at Intermountain values variation based on the patient, not the physician.

I guess my friend’s experience is one example of the Law of Unintended Consequences.  I think the rigid approach being employed in that doctor’s hospital and elsewhere is the result of little or no training in clinical process improvement in medical schools and in graduate medical education. We have often been told by residents, during our Telluride patient safety workshops, that they get more exposure to matters of clinical process improvement and high reliability systems in four days than in their seven previous years of medical training.

Brent expressed hope back in 2011:

Brent is optimistic because he has seen this philosophy of learning how to improve patient care extend to more and more doctors and hospitals around the country.  He views it as providing the answer to the rising cost of care, and he is excited about the potential.  He concludes that this is a “glorious time” to be in medicine because it is the “first time in 100 years” that doctors have a chance to institute fundamental change in the practice of medicine.

Thus far, such change only exists in certain islands of excellence, and it clearly takes energy and thoughtfulness to sustain it even in those places.

Paul Levy is the former CEO of BIDMC and blogs at Not Running a Hospital, where an earlier version of this post appeared. 

Categories: Uncategorized

Tagged as:

5 replies »

  1. I think of protocols and algorithms (in most circumstances) as where the thinking should start, not end. It is a way to frame the approach based on what is most likely to work well for the largest number nearest the center of the distribution.

    The protocol should be based on evidence, the strength and source of the evidence should be known, its applicability to the particular patient should be considered, and patient context (values, preferences…) should be considered.

    One could create a protocol for dressing to go outdoors where I live in Maine, based on reliable long term weather data. The average annual high is 54 and the average low is 36, but dressing for this range based on protocol would be uncomfortable in much of July and dangerous in most of January. Even if one looks at month-specific data, the range leaves considerable room for misery. And that is without talking about outliers.

    I think we should start by identifying what the evidence-based protocol would suggest, then carefully consider how and why we might need to alter our approach based on patient-variability (not based on our habits), and document where, how and why we altered the protocol to meet what specific circumstances.

  2. We always need to realize (as I am sure you do) that for each patient n=1. We are not doing science or statistics; we are USING science and statistics to make sure we are doing the best care we can. Guidelines and protocols are condemned too quickly by physicians, no doubt. We are afraid to find out where we are deviating from the norm, when we should be grateful to know that reality. All physicians have inside their own head protocols as to what “normal” looks like for certain kinds of patients. This is called good care. On the other hand, assuming that deviating from normal is always indicative of bad care is also a problem. Docs need to know when they are outliers. We need to know when we are outside of the norm, but only for information. Being punitive in this case is not valuable in an environment where docs feel targeted (and where they are the key factor in controlling cost). In general, we docs should always hold ourselves to higher standard than that which any protocol puts us under. That is the definition of being a good doctor.

  3. Perhaps part of the resistance to protocols in medicine is not just the loss of autonomy but the addition of multiple extra steps for even the simplest processes. Physicians are drowning in their computer screens already. Often a new protocol introduces multiple extra checklists, red flags, and automatic triggers in the EHR that must be addressed before meds can be ordered or fluids given.

    Add in the fact that now many protocols are triggered automatically by EHRs and often enroll the wrong patients. Every single one of those patients the physician now has to spend time documenting why the protocol did not apply. That is time lost from patient care.

    I think protocols clearly can improve patient safety and should be done. But there should be equal emphasis at studying how they will impact the efficiency of an entire department, not just the individual patient they are applied to. If protocols are designed that make patient care safer and at the same time improve the efficiency of an entire department, I think you will find physicians would be more than willing to implement them.

  4. We wouldn’t need protocols if we would consistently practice evidence-based medicine. How often do our colleagues treat routine otitis externa with PO antibiotics? How often do we prescribe antibiotics for viral sinusitis/bronchitis? How many people are addicted to narcotics despite the fact that there is no good evidence to support their use in chronic, non-cancer pain? How many unnecessary CT scans are done daily in our ER’s? How many people have died because we failed to recognize the early signs of sepsis? We fail to police ourselves but then we complain when the dreaded protocols are forced on us.

  5. Good discussion. Thanks. I wonder if we know enough to have very many solid protocols. The science is changing so fast it seems. Look at the variation in protocols themselves. Also, I wish folks would begin thinking about getting the correct diagnosis a little more. I want a few patients to have acute intermittent porphyria and not appendicitis and a few to have Mediterranean Fever instead of migratory arthritis, etc. I hope artificial intelligence can move along faster and help us here.