Ambulatory Glucose Management Workgroup Minutes

January 27, 2026

3:00pm – 4:00pm EST

Zoom

Troy Wildes, University of Nebraska

Diego Bauza, Weill Cornell

Patrick Henson, Vanderbilt

Mara Bollini, Washington University

Tony Edelman, University of Michigan

Josh Goldblatt, Henry Ford Jackson

Megan Charette, MPOG

Tariq Esmail, University Health Network

Nirav Shah, MPOG

Mason Smith, MyMichigan Health

Kate Buehler, MPOG

Brandon Yee, WashU

Meridith Wade, MPOG

Xan Abess, Dartmouth

Paige Georgiadis, University of Vermont

Bethany Pennington, WashU

Meeting Start: 1500

Purpose of the Workgroup

The group is developing perioperative ambulatory hyperglycemia measures for MPOG, focusing on:

• Perioperative glycemic management in outpatients

• Emphasis on hyperglycemia, not solely patients with diagnosis of diabetes

• Practical, meaningful processes suited to short-duration, fast-paced ambulatory workflows

December Meeting Recap

Core Agreements on Hyperglycemia Treatment measure

• Population: Patients presenting with hyperglycemia, regardless of diabetes diagnosis.

• Thresholds:

o GLU15: BG > 250 mg/dL

o GLU15b: BG > 180 mg/dL

• Measure Focus: Treatment timeliness—not insulin dosing quantity.

December Decisions Reconfirmed

• Build two measures: GLU15 and GLU15b.

• Treatment window: Draft at 60 minutes.

• GLU15b to remain informational initially (no threshold).

Review of Draft Specifications: GLU-15 Draft Spec - Google Docs

Measure Intent

Percentage of adult outpatient procedures where a high glucose value receives insulin treatment within

60 minutes.

Time Period: Preop Start → PACU End

Inclusions: All adult outpatient procedures with qualifying hyperglycemia.

• GLU15 (250 mg/dL): Process measure with a performance target (initially 90%).

• GLU15b (180 mg/dL): Informational submeasure, no pass/fail threshold.

Treatment Window (60 vs 90 minutes)

Concerns raised (Vanderbilt):

• 60 minutes may be unrealistic in short, high-workflow cases.

• A “fail” at 61 minutes may be unfair.

Counterpoint (Henry Ford):

• Their protocol targets 30-minute treatment.

Consensus:

Keep 60 minutes in the draft, but MPOG will analyze performance at 60 vs 90 minutes and bring data

back to the group to guide refinements.

Action: MPOG QI/Data team to conduct timing analysis.

Definition of Treatment

Counts as treatment:

• IV insulin infusion (active during the window)

• SQ insulin administration

Does NOT count:

• Glucose recheck alone

Rationale: GLU15/15b track treatment only. Rechecks handled in separate measures.

Handling Clinical Edge Cases

Example: Patients self-administer insulin at home before arrival.

Group conclusion:

Some cases will be flagged despite appropriate care. Thresholds are not 100%, and flagged cases

support review—not punitive interpretation.

Measure Logic Safeguards

To avoid encouraging overtreatment:

• Values within 60 minutes of PACU discharge are excluded.

• SubQ dosing window: High values within 120 minutes of SQ insulin are not counted as new

events.

• IV insulin infusion: If active within 60 minutes, value passes.

• Repeated values: Designed to avoid repeated penalties.

The group agreed these guardrails appropriately reduce pressure for insulin stacking.

Hyperglycemia Assessment Measures

Rationale

Separate, scenario-based assessment measures are preferred over a single complex measure. This

aligns with MPOG’s cardiac antibiotic bundle model.

Candidate Assessment Components

Likely standalone measures:

1. Preoperative glucose check for diabetic patients

2. Recheck after perioperative BG > 180 mg/dL

3. Glucose check after IV insulin within 90 minutes

4. Glucose check after SQ insulin within 120–150 minutes

5. (Tentative / debated) PACU glucose check after intraoperative insulin

Key Discussion Points

Clinical Meaningfulness

• SQ insulin may not show effect quickly—premature rechecks not actionable.

• Rechecks should be tied to time since insulin rather than location (e.g., “in PACU”).

Workflow Variation

• In many ASCs, rechecks naturally occur preop + PACU, not intraoperatively.

• Expectations must avoid labeling appropriate ASC workflows as failures.

Measure Design

• Avoid overprescribing local protocols.

• Granular components allow sites to focus on processes aligned with their workflow.

Emerging Consensus

• Build separate measures for each assessment scenario.

• Also build a composite/bundle measure representing “all applicable assessment components

met.”

• Provide dashboards showing each component + composite.

Action: MPOG QI team to draft:

• Individual assessment measure specifications (including GLU16 for diabetic preop assessment)

• Composite bundle concept and dashboards

Diabetes Phenotype Development

Accurate identification of diabetic patients is required.

Proposed multisource phenotype inputs:

• Problem list

• Preoperative medical history

• ICD-10 codes (outpatient + inpatient)

Action:

MPOG Phenotyping team to build phenotype and circulate test cases for validation (volunteers: Tariq,

Josh, others).

Next Steps

• Distribute finalized minutes to the workgroup.

Measure Specification Work

• Update GLU15/15b specs with measure status distinctions.

• Draft assessment measures + composite bundle.

• Incorporate timing clarifications (e.g., tie rechecks to time since insulin).

Data Analysis

• Compare treatment performance at 60 vs 90 minutes.

• Test feasibility of recheck timing rules using retrospective data.

• Programmer assignment after feedback.

Future Meeting

• Leadership will determine if another live meeting is needed following asynchronous feedback

and preliminary data review.

Appendix A — Full Transcript

Nirav J. Shah (MPOG)

Quick recap: this workgroup is looking at glycemic management—specifically hyperglycemia—in the

ambulatory space. We had a fantastic discussion last time, so thank you to those who joined and

participated. It was very thoughtful and well represented.

A few themes from December:

• We discussed whether to focus only on patients with diabetes versus patients with

hyperglycemia. It became clear that we want to focus on patients with hyperglycemia, whether

or not they carry a diagnosis of diabetes.

• We talked about unique characteristics of the outpatient space compared with inpatient or

admitted patients: shorter case duration, shorter time with us overall, and the predominant use

of subcutaneous insulin rather than IV insulin in most centers.

• We debated what constitutes hyperglycemia in the ambulatory space—180 vs 250 mg/dL. We

couldn’t pick one, so we decided to use both thresholds, which we felt was reasonable. That’s

why we have GLU-15 and GLU-15b.

• There was consensus around emphasizing timely treatment rather than dosing specifics. We’re

not focusing on how much insulin is given, because that is nuanced, protocol-based, and difficult

for MPOG to encode; instead, we focus on whether treatment was given in a timely way and

whether blood glucose is appropriately checked.

We agreed to build two measures:

• GLU-15: Threshold 250 mg/dL (13.9 mmol/L), treatment within 60 minutes.

• GLU-15b: Corollary measure for blood glucose >180 mg/dL, also with treatment within 60

minutes. We can flip which one we call 15 and 15b, but we decided to build both.

Any questions or comments on the December meeting before we move into a more detailed review of

the measure specifications?

If anyone has comments or feedback, feel free to unmute and chime in or raise your hand.

(Brief pause; no comments.)

Alright, let me share the specification, and we’ll go back to some of the additional considerations as

well.

Overview of GLU-15 / GLU-15b Specs

Nirav J. Shah (MPOG)

This is a measure specification. For those familiar with MPOG measures, this should look like a familiar

format.

GLU15 and GLU15b measure the percentage of patients undergoing an outpatient procedure with blood

glucose greater than 250 mg/dL (for GLU15) or 180 mg/dL (for GLU15b) who are treated within 60

minutes with insulin. The target performance threshold is 90%, which is common for our measures. It is

consensus driven, not data driven, and we can adjust it after we review baseline performance.15 and

GLU15b measure the percentage of patients undergoing an outpatient procedure with 15) or 180 mg/dL

(for GLU15b) who are treated within 60 minutes with insulin. The target performance threshold is 90%,

which is common for our measures. It is consensus driven, not data driven, and we can adjust it after we

review baseline performance. GLU-15 and GLU-15b measure the percentage of patients undergoing an

outpatient procedure with GLU-15) or 180 mg/dL (for GLU-15b) who are treated within 60 minutes with

insulin. The target performance threshold is 90%, which is common for our measures. It is

consensus-driven, not data-driven, and we can adjust it after we review baseline performance.

The measure time period is Preop Start to PACU End, so not just intraoperative; it includes pre-op

holding and PACU.

Tony Edelman (University of Michigan / ASPIRE–MPOG)

When we go back to the description, I thought last time we discussed that GLU-15 would be a process

measure with a goal ≥90%, but GLU-15b would perhaps be informational initially, without a threshold,

because it’s the more aggressive target.

Nirav J. Shah (MPOG): That’s a good point. I do remember that. We’ll note that distinction—

GLU-15 as a process measure with a threshold, GLU-15b initially informational.

Debate on Treatment Window: 60 vs 90 Minutes

• Patrick Henson (Vanderbilt University)

I’ve been thinking about the timing threshold. I feel like 60 minutes is a hard requirement. I

think 90 minutes may be more appropriate. I wanted to make sure this was open for discussion.

Sixty is a good target, but missing it by one or two minutes would cause a fail, which might be

problematic.

o Nirav J. Shah (MPOG): That’s a fair point. We can go in either direction. Sometimes as

long as measure performance is interpreted in context—for example, if insulin is given

at 65 minutes and the case was hectic—that could still represent reasonable care.

• Patrick Henson (Vanderbilt University)

One thing I often miss—not being the bedside provider—is how much work goes into starting a

quick case. Those tasks often take priority. I’ve heard anecdotally that 60 minutes can be tough

to meet, so I’d hate to have failures solely based on that timing.

o Nirav J. Shah (MPOG): That’s a really good point.

• Josh Goldblatt (Henry Ford Health)

At Henry Ford, we have a single protocol for inpatient and outpatient/ambulatory. Our

expectation is to treat a high glucose within 30 minutes. After more time, the glucose value may

no longer accurately represent the patient’s current state—it could have gone up or down. So

our protocol is actually more aggressive: treat within 30 minutes.

• Nirav J. Shah (MPOG): So we’re hearing arguments both for a longer window (90 minutes) and

for even shorter (30 minutes). One thing we could do, which we didn’t do when we built the

original measures, is test performance at different time frames—60 and 90 minutes—and see

what types of cases fall between 60 and 90 minutes. Then we can share that back for

asynchronous discussion to see if the group still feels strongly about modifying the window.

• Patrick Henson (Vanderbilt University): I’d support whatever the group feels is accurate. I just

think switching from 90 to 60 will make this harder to achieve. But we do tackle hard things.

• Nirav J. Shah (MPOG): We also think about “minimum acceptable” versus “gold standard.”

Traditionally, MPOG has started at a threshold that feels broadly acceptable and then ratcheted

higher over time—as we did with sustainability measures, and early glucose measures. I’d be

comfortable doing something similar here. If it’s okay with everyone, we’ll start with 60 minutes

in the preliminary analysis and also test other thresholds (like 90 minutes). We can then see

whether flagged cases in the 60–90-minute band seem reasonable or unfairly flagged.

• Nirav J. Shah (MPOG): Diego, you asked whether treatment includes rechecking. That’s a great

question. For this measure, treatment does not include rechecking; we’re purely looking at

whether a patient with hyperglycemia was treated with insulin.

“Treatment” vs “Recheck” and Edge Cases

• Branden Yee (WashU): I’m sorry I missed the prior meeting. I do a lot of outpatient-based

procedures. Sometimes highly educated patients discover that their sugars are, say, 250 at

home and give themselves insulin before coming to the hospital. When we check while starting

the IV, they might still be above threshold but trending down. Based on our current outline, this

would be a fallout, even though we did the right thing by not treating again because they’re

already down-trending. How do we account for that?

• Nirav J. Shah (MPOG): That’s a really good point, and it’s characteristic of almost all our

measures. There will be cases where the measure definition cannot capture all clinical nuance,

such as self administered insulin prior to arrival that isn’t reliably documented in a structured

fashion.-administered insulin prior to arrival that isn’t reliably documented in a structured

fashion.

That’s one reason we changed terminology years ago from “failed case” to “flagged case.” The

measure might flag a patient whose care was entirely appropriate, but that flag becomes an

opportunity for review, not a judgment. Also, we don’t set thresholds at 100%; there is room for

such exceptions.

• Troy Wildes (University of Nebraska): I hope I’m not jumping ahead, but I have a related

question about common workflows. For many shorter ambulatory procedures, it may be

appropriate and feasible to treat patients once in their perioperative period. Because of the

duration of effect of sub-Q insulin, we may not treat again even if we recheck. I’m concerned

that we might discourage rechecks if we always expect repeat treatment. How will the measure

handle this?

• Nirav J. Shah (MPOG): We definitely want to avoid incentivizing the wrong behavior, not just

here but with all measures. How do you see this measure possibly encouraging unintended

behavior?

• Troy Wildes (University of Nebraska)

: Broadly, I’d suggest: if we have an ambulatory patient

who is hyperglycemic and they receive one appropriate perioperative treatment, that might be

a reasonable “success.” We might think about a measure focused on “did the hyperglycemic

patient receive treatment at some point in association with their procedure” rather than linking

each elevated value to a separate treatment. That might be a lower, reasonable target.

o Nirav J. Shah (MPOG): I’d like to hear what others think.

• Patrick Henson (Vanderbilt University): We have to be careful not to prescribe institutional

policy through these measures. We already experience some of that where I am—what we can

measure and report is not always the full story. Local judgment still needs to govern decisions

like whether to re-dose.

• Troy Wildes (University of Nebraska): Exactly; it’s different from inpatient procedures where

you’re on an infusion with frequent checks and titrations. In outpatient settings, a single early

sub-Q dose may be the best you can reasonably do.

• Tony Edelman (University of Michigan / ASPIRE–MPOG)

If we zoom out, there are three main steps:

1. If a patient is diabetic or at risk: Did we check a glucose?

2. If we checked and they were hyperglycemic: Did we treat?

3. If we treated: Did we recheck?

In complex inpatient cases, we do all three repeatedly. In the ambulatory setting, GLU15/15b is

focusing only on step 2: if you checked and the patient is hyperglycemic, did you treat with

insulin? The “bookends” (initial checks and rechecks) aren’t captured here.-15/15b is focusing

only on step 2: if you checked and the patient is hyperglycemic, did you treat with insulin? The

“bookends” (initial checks and rechecks) aren’t captured here.

We saw this at Michigan when we hardwired our ambulatory perioperative glucose protocols.

Cataract patients, for example, might have a total perioperative time of 40 minutes. They’d be

hyperglycemic, be treated (so they pass GLU15), but our internal protocol also required a follow-

up glucose and potentially a second dose if still high. That recheck requirement was causing

throughput issues in PACU. The fact that GLU15 only tracks “if hyperglycemic, did you receive

insulin?” gives flexibility: it doesn’t require rechecks or multiple doses.-wired our ambulatory

perioperative glucose protocols. Cataract patients, for example, might have a total perioperative

time of 40 minutes. They’d be hyperglycemic, be treated (so they pass GLU-15), but our internal

protocol also required a follow-up glucose and potentially a second dose if still high. That

recheck requirement was causing throughput issues in PACU. The fact that GLU-15 only tracks

“if hyperglycemic, did you receive insulin?” gives flexibility: it doesn’t require rechecks or

multiple doses.

• Josh Goldblatt (Henry Ford Health): For my team, consistency in the structure of measures is

important. They don’t want us to reinvent how we calculate success each time. Having a similar

build to existing measures makes it easier to educate and interpret results.

• Nirav J. Shah (MPOG): That makes sense. Josh, can you speak a bit more about your protocols at

Henry Ford? Do they distinguish between pre-treatment, recheck, and re-treatment, or is it

more of a “once per patient” approach?

• Josh Goldblatt (Henry Ford Health): We have multiple pathways, and you’re on a pathway until

you leave. Non-diabetic patients are on a lower-dose scale; those with kidney disease or other

issues are adjusted accordingly. Testing intervals depend on whether you’re non-diabetic with

normal blood sugar versus type 1 or type 2 diabetic.

We also have different intervals in different phases of care. In preop, if you’re held for a while,

we don’t need the same testing frequency as under anesthesia. It’s not a simple algorithm,

which is why we just said “this is the algorithm everywhere.”-op, if you’re held for a while, we

don’t need the same testing frequency as under anesthesia. It’s not a simple algorithm, which is

why we just said “this is the algorithm everywhere.”

• Nirav J. Shah (MPOG): That’s helpful. My perspective is that we may lose some opportunities for

case review if we don’t capture subsequent episodes of hyperglycemia in longer outpatient

cases (2–4 hours). Not because clinicians are doing the wrong thing, but because we won’t see

those events as flags. We’ve built some measure details to account for timing. For example, if

you have a high blood sugar right before discharge and no time to treat, it’s excluded. So the

measure isn’t blind to timing.

We may decide over time to modify measures in either direction—more stringent or more

relaxed. For now, if in general we’re asking clinicians to treat a high glucose in preop, recheck it

in an appropriate time frame, and then consider another dose if still high, I’d prefer to keep that

concept in our measures and let the timing details protect against unfair flagging.-op, recheck it

in an appropriate time frame, and then consider another dose if still high, I’d prefer to

• Tony Edelman (University of Michigan / ASPIRE–MPOG): If we scroll down to the measure

details, I think that’s where some of the concern is addressed.

Measure Logic Details (Treatment Windows, Exclusions)

• Nirav J. Shah (MPOG): In the details, we take each high glucose value and decide how to classify

it.

• If the value occurs within 60 minutes of measure end (PACU end), it is excluded, because

there isn’t enough time to treat and see an effect.

• If insulin was given within 60 minutes after that high value, that’s a pass.

• If an IV insulin infusion is administered and is running within 60 minutes of a high value, that

is also considered passing.

• For subQ insulin, we adjust timing: If you have a high glucose value within 120 minutes of a

prior SQ insulin dose, we exclude that value, acknowledging the expected pharmacokinetics.

We also exclude certain “follow up” values: for example, if there is a repeated high value within

90 minutes of an earlier high value >180 mg/dL, we may not count that as a new failure,

depending on context; we still expect insulin to have been given within 60 minutes of the initial

elevated glucose.-up” values: for example, if there is a repeated high value

• Tony Edelman (University of Michigan / ASPIRE–MPOG): I think points 1, 4, and 5 in the

detailed logic are important. Points 4 and 5 limit the need to stack doses. You can recheck

without being penalized if you don’t give a second dose yet. There’s also no negative

consequence to checking in PACU when the patient is nearly ready for discharge.

• Patrick Henson (Vanderbilt University): Since you treated, wouldn’t you already have

success? Oh, I see—this addresses cases where a high value comes after a dose was already

given and is still within the effective window.

o Nirav J. Shah (MPOG): Exactly. If you have a high value that occurs within the

pharmacologic window of a prior dose, the logic can exclude it rather than treating

it as an independent failure. We’re at the point where I’d like to pivot to the

assessment component, because we expect a robust discussion there. We will

circulate these GLU15/15b specs to this group after the call so people can review

and comment before we start coding. GLU-15/15b specs to this group after the call

so people can review and comment before we start coding.

Hyperglycemia Assessment Measures (Outpatient)

Rationale and Proposed Scenarios

Nirav J. Shah (MPOG): The other half of ambulatory glycemic management is when and in what

circumstances we should be checking blood glucose.

Over the last few years—through discussions with some of you, other colleagues across MPOG, and

internal discussions—we’ve reached a couple of conclusions:

1. There are multiple distinct situations in which we should be checking blood glucose.

2. In early glucose measures, we combined treatment and blood glucose checking into a single

metric (e.g., older versions of GLU-01 and GLU-03). Over time, those evolved into GLU13 and

GLU-14, reflecting our learning.-1 and GLU-3). Over time, those evolved into GLU-13 and

GLU-14, reflecting our learning.

For these ambulatory measures, we propose to separate treatment from blood glucose checking.

Because there are several scenarios where checking is important, we think separate assessment

measures make sense.

We’ve identified several potential scenarios:

1. History of diabetes:

o If a patient has a documented history of diabetes, a preoperative glucose should be

checked preoperatively.

2. Hyperglycemia:

o If a patient has blood glucose greater than 180 mg/dL (or possibly 250 mg/dL)

perioperatively, the glucose should be rechecked within 90 minutes.

3. IV insulin infusion:

o If an IV insulin infusion is started, glucose should be checked within 90 minutes of

infusion start.

4. Sub-Q insulin:

o If SQ insulin is administered, glucose should be checked within 120 minutes, or possibly

120–150 minutes, to account for longer onset and peak.

5. PACU check after intraoperative insulin:

o An earlier idea was: if insulin is administered intraoperatively, there should be a PACU

glucose checked. We’ve already flagged that for reconsideration and, in the slide, it’s

crossed out.

I’d like to pause and get thoughts: Are these the right scenarios? Is the timing appropriate? Are we

missing any? Are any inappropriate?

Concerns About PACU Check After Insulin and Timing

• Tony Edelman (University of Michigan / ASPIRE–MPOG)

Looking at the bullet about “if insulin was administered intraoperatively, a PACU glucose was

checked,” my concern—especially in the ambulatory space—is that if we’re giving Lantus or

regular insulin over a short perioperative period, it may not make clinical sense to recheck so

quickly.

If we check soon after giving a subQ dose and it’s still elevated, we may not do anything

clinically, because we’re still waiting for the peak effect. Requiring that check could generate

values that aren’t actionable.-Q dose and it’s still elevated, we may not do anything clinically,

because we’re still waiting for the peak effect. Requiring that check could generate values that

aren’t actionable.

We might consider adding guardrails around timing—for example, only requiring a PACU check

if a certain time has elapsed since insulin dosing.

Nirav J. Shah (MPOG): That’s a good point. One way this would work is if the insulin was given far

enough before discharge—for example, if the SQ insulin was given more than 120 or 150 minutes before

PACU end—then we’d expect a recheck. If the patient is discharged sooner, perhaps not.

Tony Edelman (University of Michigan / ASPIRE–MPOG): Yes, requiring a check after an appropriate

interval (rather than immediately in PACU) would be more clinically meaningful.

Nirav J. Shah (MPOG): So just being in PACU shouldn’t automatically mean they get a blood glucose

check; the expectation should depend on time since insulin.

Henry Ford Experience with Recheck Timing

• Josh Goldblatt (Henry Ford Health): We talked a lot about recheck timing when we revised our

protocol. We wanted to align with existing MPOG metrics, which were built around treatment.

We liked the idea of a test before maximal peak effect, to stay on top of patients who are labile

and may need another dose. Waiting longer than two hours didn’t seem appropriate. For

inpatient treatment metrics (e.g., GLU10), if we recheck within two hours after insulin, that

value can be excluded from retreatment expectations. Our orders for insulin and glucose testing

are not linked directly in the EHR; they each have their own cadence, which complicates tying

them together.-10), if we recheck within two hours after insulin, that value can be excluded

from re-treatment expectations. Our orders for insulin and glucose testing are

In our protocol, we recheck at 90 to 120 minutes, which gives us:

1. A safety check before full peak effect.

2. A check time that is still less than the 120-minute window in the metric.

• Nirav J. Shah (MPOG): For these proposals—e.g., “recheck within 90 minutes if glucose >180”—

would you modify any of them?

o Josh Goldblatt (Henry Ford Health): If our expectation is that a glucose over 180 is

treated, we don’t necessarily need a measure to tell us whether it was rechecked; that’s

already part of the treatment logic.

If we want to know whether we’re doing ongoing monitoring, that should be tied to the

insulin administration, not just the high value. A metric tied to a check two hours after

insulin would show whether we’re monitoring promptly without contradicting the

treatment algorithm.

• Nirav J. Shah (MPOG): Is any of this currently contradicting the treatment measure?

o Josh Goldblatt (Henry Ford Health): Bullet four might. As soon as we’re outside the two

hour timeframe, a high glucose could count as a new event requiring treatment, even

though clinically it might still be within the expected effect window, depending on the

insulin and patient.-hour timeframe, a high glucose could count as a new event

requiring treatment, even though clinically it might still be within the expected effect

window, depending on the insulin and patient.

• Tony Edelman (University of Michigan / ASPIRE–MPOG): Point two is also potentially

problematic: if you require a recheck within 90 minutes of the high value, and you treat at

minute 58, you only have 32 minutes left to recheck. That might not be clinically sensible.

Maybe that timing should be tied to the insulin dose rather than the initial high value.

o Nirav J. Shah (MPOG): Right now bullet two is values pecific. We could consider

“resetting the clock” at the time of insulin administration, rather than anchoring it

entirely to the initial high value.-specific. We could consider “resetting the clock” at the

time of insulin administration, rather than anchoring it entirely to the initial high value.

• Patrick Henson (Vanderbilt University): Do bullets 3 and 4 override bullet 2? If not, bullet 2

might still create an unrealistic expectation.

• Nirav J. Shah (MPOG): At present they don’t override bullet 2. We can change that in the spec,

so if insulin is given, the requirement for recheck timing is tied to the insulin, not the original

hyperglycemia timestamp.

Complexity, Site Variation, and Bundle vs Single Measure

• Tariq Esmail (UHN Toronto): How do you evaluate these once the measure launches to decide

whether they’re actually useful? Our preop nurses might check a glucose when someone arrives

at the ASC. An anesthesiologist might not see the patient for 30 minutes, and surgery might

start 45 minutes later. We don’t typically check intraoperatively at our ASC (no routine intraop

POC testing), just preop and PACU unless there’s an arterial line. I’m worried that some cases

might be flagged as issues even though the most appropriate standard of care is: preop check,

then postop check and treatment as needed. How will we sift out these situations?-op nurses

might check a glucose when someone arrives at the ASC. An anesthesiologist might not see the

patient for 30 minutes, and surgery might start 45 minutes later. We don’t typically check

intraoperatively at our ASC (no routine intraop POC testing), just pre-op and PACU unless there’s

an arterial line.-op check, then post-op check and treatment as needed. How will we sift out

these situations?

• Nirav J. Shah (MPOG): We’ve thought about that. One option is to mirror what we do in the

cardiac antibiotic bundle: Separate measures for timing, selection, and redosing, plus a

composite bundle that passes only when all components are met. We could do something

similar here: build out separate measures for each assessment scenario (pre-op check in

diabetics, recheck after high glucose, monitoring after insulin, etc.), then compose a bundle

measure.

• Kate Buehler (MPOG): I agree. This would be better suited as separate measures with a

composite for overall hyperglycemia assessment. Trying to pack four or five assessments into a

single measure will create numerous caveats in the code and make it very hard to debug or

interpret.

For example, “was a preop glucose checked in diabetics” is a very different process from “did

you recheck after treatment.” I can already think of several caveats just for one bullet.-op

glucose checked in diabetics” is a very different process from “did you recheck after treatment.”

I can already think of several caveats just for one bullet.

If we have separate measures, sites can pick and choose which align with their policies. If they

like all of them, they can look at the composite. A dashboard could show the components side

by side and the bundle on the same screen.-by-side and the bundle on the same screen.

• Tariq Esmail (UHN Toronto): I like that, especially for “if they have a history of diabetes, was

glucose checked preoperatively?” That’s a distinct process from what the anesthesiologist

does intraoperatively. In our ASC, there are fewer handovers than at our tertiary centers—

you often stay with your room and your trainee. I see less value in the “rechecked within 90

minutes” piece itself if surgery is short and I wouldn’t recheck intraoperatively anyway. It

might be more useful to know “was it rechecked before anesthesia end or before

discharge,” especially if that recheck is handled by PACU nurses.

• Nirav J. Shah (MPOG): That’s a great point. We’re also not limiting this to ASC cases; it

includes any outpatient procedure, though ASC is one important subset.

• Tariq Esmail (UHN Toronto): If we can toggle individual measure components and the

composite on and off, that would help. Sites could align measures with their own workflows.

• Nirav J. Shah (MPOG): : We wouldn’t build the composite “on the fly” at each site, but we

could configure it such that a site can look at:

• Each component individually, and

• The composite “all components met” status.

It would also make it easier to refine each individual measure based on how they

perform and how sites react.

Kate Buehler (MPOG): We could potentially put the bundle metrics on a single dashboard so you can

see them side-by-side. You’d quickly see: “Our composite is low, but it’s driven by this specific measure

we don’t agree with or don’t use,” or “Here’s the component that’s truly our area for

improvement.”-by-side. You’d quickly see: “Our composite is low, but it’s driven by this specific measure

we don’t agree with or don’t use,” or “Here’s the component that’s truly our area for improvement.”

Patrick Henson (Vanderbilt University)

We’re just starting to build ambulatory glycemic management capacity. Any granular data will help us. I

like having a composite, but we’ll be very interested in how well we:

• Check diabetic patients pre-operatively,

• Recheck when indicated, and

• Treat appropriately.

The more granular we can be, the more we can target practice improvement.

Diabetes Phenotype and Data Sources

Nirav J. Shah (MPOG): Another point: we’re not 100% sure how accurate our diabetes diagnosis data is

in MPOG for the purpose of quality measures. We have used diagnosis codes for research, but not as

much in quality measures.

We will need to test whether we can reliably identify diabetics using discharge diagnoses, problem lists,

and pre-op documentation.

Tariq Esmail (UHN Toronto): Do you have the ability to look at it from multiple sources? At our site,

diabetes may be documented in:

• The pre-admission clinic note or history,

• The problem list.

We don’t rely on billing codes or discharge diagnoses in the same way.

Nirav J. Shah (MPOG): Yes. We’d likely build a diabetes phenotype that uses:

• Problem list entries,

• Pre-op medical history,

• Diagnosis codes.

Kate Buehler (MPOG): We have existing comorbidity phenotypes that rely on ICD10 codes. We’d

probably bundle those with conceptbased identifiers to create a diabetes phenotype. It won’t be

perfect—we’ll need sites’ help to refine it.-10 codes. We’d probably bundle those with concept-based

identifiers to create a diabetes phenotype. It won’t be perfect—we’ll need sites’ help to refine it.

Tariq Esmail (UHN Toronto): The unintended consequence may be that we all document

diabetes more accurately, which is actually good.

Kate Buehler (MPOG): Exactly. We saw that with smoking. Once we introduced smoking

metrics, smoking documentation improved substantially.

Composite Approach and Dashboards

Nirav J. Shah (MPOG): It sounds like there’s broad interest in:

• Separate assessment measures, and

• A composite to summarize overall outpatient hyperglycemia assessment.

We will likely start by building each component measure and then the composite.

As we build and spec these measures at the Coordinating Center, we’ll learn about potential pitfalls and

share them with the group. If we encounter a critical issue that threatens measure validity, we may

reconvene to decide whether to move forward despite known limitations or adjust course.

Kate Buehler (MPOG): We can also design dashboards that show the bundle metrics together, making it

easy to see which element has the lowest performance and should be prioritized for quality

improvement.

Measure Development Process, Validation, and Next Steps

Tariq Esmail (UHN Toronto): A practical question: when you’re developing new measures like this, what

happens next? Do you test them with data only you can see, or do you pilot them at a single site like

Henry Ford? How does the process work?

Nirav J. Shah (MPOG):

We do a lot of the initial work at the Coordinating Center. Once we reach a certain point, we rely on

individual sites to help with validation, especially chartlevel validation.-level validation.

Kate Buehler (MPOG): It mostly depends on willing volunteers. If I send you cases and you’re willing to

validate them, I don’t particularly care what site you’re at—I’m just thrilled you’re helping. That’s how

we’ve done it to date: friends and family at sites around the country (and now internationally) review

cases and help us refine the measures.

Summary of Next Steps and Meeting Close

Nirav J. Shah (MPOG):

For next steps:

• We will share the minutes from this meeting.

• We have preliminary specs that need to be fleshed out and then sent to this group.

• Based on the feedback we receive on those specs, we’ll decide whether we need another live

meeting or if we can begin assigning measures to programmers and start coding.