Definition

Correlation between HbA1c level and mean plasma glucose
Hemoglobin A1c	Mean plasma glucose
%	mg/dl	mmol/L
6	135	7.5
7	170	9.5
8	205	11.5
9	240	13.5
10	275	15.5
11	310	17.5
12	345	19.5

“Glycated hemoglobin (GHb), also commonly referred to as glycosylated hemoglobin, glycohemoglobin, HbA1c ... [is] a series of stable minor hemoglobin components formed slowly and nonenzymatically from hemoglobin and glucose. The rate of formation of GHb is directly proportional to the ambient glucose concentration. Because erythrocytes are freely permeable to glucose, the level of GHb in a blood sample provides a glycemic history of the previous 120 days, the average erythrocyte lifespan.”

David E. Goldstein, MD, Randie R. Little, PHD, et. al.. Tests of Glycemia in Diabetes. Diabetes Care, 27:7, (July 2004), pp.1765, 1767.

Clinicians used to think that elevated blood sugar caused disease slowly with years of cumulative exposure to hyperglycemia leading to poor outcomes. Given this time scale—months to years—the choice of hemoglobin A1c as an indicator of glucose control was appropriate. Hemoglobin A1c correlates with blood glucose levels in the three months preceding its measurement.

Short-term hyperglycemia occurring during a hospitalization for another illness (say, pneumonia or a myocardial infarction) was considered of minor importance. Now several studies of acutely ill patients suggest that short-term hyperglycemia in any hospitalized patient, diabetic or not, is an independent risk factor for poor inpatient outcomes, including mortality. In fact strict control of blood sugar in these patients reduced in-hospital mortality from 11 to 7 percent. Those results coupled with these statistics—twenty million Americans with diabetes and an inpatient diabetes prevalence of up to 25%—account for the growing interest nationwide in improving inpatient glucose management.

To improve management one needs an indicator that correlates with blood glucose levels during the short interval of an inpatient stay, usually a few days. Hemoglobin A1c responds too slowly to measure glucose fluctuations over days. Instead studies of inpatient glycemia have correlated outcomes with glucose measurements: both single glucose values (admission, random, fasting, and maximal glucose) and mean glucose values (admission, first 24 hours of hospitalization, fasting, post-operative, and overall mean). However the lack of a standard hampers comparison whether between studies, institutions, or quality improvement projects.

A candidate measure

Philip A. Goldberg MD, Janis E. Bozzo RN,MSN, Prem G. Thomas MD, Melinda M. Mesmer MD, Olga Sakharova MD, Martha J. Radford MD, and Silvio E. Inzucchi, MD. "Glucometrics" – Assessing the Quality of Inpatient Glucose Management. Diab Technol Ther, 8:5, (October 2006), 560-569.

We recently proposed a composite measure of inpatient glycemic control with three components: glycemic exposure, efficacy of control, and rate of adverse events. In addition, we applied the measure to a cohort of patients using three different 'units of analysis.'

Units of analysis

All the units of analysis involve a patient, a time interval, and a glucose level (a single measurement or a mean).

One unit, the individual glucose sample from a patient, gives a look at the glucose level in the time interval between two other samples: the one right before and the one right after. This unit generally covers the shortest time interval, usually a few hours. Call it the patient‑sample.

A second unit, the mean of all glucose samples taken from the patient during a stay (on a single ward or in the hospital), covers the longest time interval, usually a few days. Call it the patient‑stay.

A third unit, the mean of all glucose samples taken from the patient during one day, is intermediate in duration. It is the only unit specifying a fixed time interval. Call this the patient‑day.

While the first two units have uses, we generally recommend the third unit. Here's why.

Consider the outpatient setting: Hemoglobin A1c correlates with glucose levels during a fixed time interval, three months. This makes possible valid comparisons of glycemic control. Without the standard time interval, a Hemoglobin A1c of 6.5 in one patient may mean good control for 4 months and in another may mean good control for 12 months. Likewise, valid comparison of glycemic control in inpatients requires the same measurement interval between one patient and another. The figure below illustrates the problem with varying intervals: three patients have the same mean glucose, but these means reflect three intervals of control.

Both the time between samples and the length of a patient's stay vary. In the intensive care unit glucose samples may be drawn hourly; in other units samples are taken every 6, 8, or 12 hours. Similarly, the length of stay varies—two, four, seven days or longer. To standardize measurement we recommend a fixed time unit of 24 hours.

Measure components

We now return to the three components of our candidate measure.

The single glucose level (patient-sample) or the mean glucose level (patient-stay, patient-day) measures glycemic exposure.

Discussions of quality of care often reference the efficacy and safety of treatments. The proportion of single glucoses (patient-sample) or mean glucoses (patient-stay or -day) within a target or goal range measures efficacy of control. Any intervention attempting to strictly control hyperglycemia may increase the rate of hypoglycemia. And marked hyperglycemia may lead to poor outcomes. These two adverse events should be monitored.

In calculating an event rate, one can count the number of adverse events per time interval and report the mean for all intervals. However, since most intervals do not have an adverse event, an alternate formulation may be useful: count the number of time intervals with at least one adverse event and report the count as a proportion of all intervals.

These are the glucometrics for a population of patients.

Example calculation with the patient-day unit

For a given inpatient cohort, each patient's hospital stay is divided into patient-days. Only days with glucose measurements are used for metrics. The mean glucose of each patient-day is calculated. Then each patient-day is tagged as within a goal range if the patient-day mean fell between 70 and 149 mg/dl, 'hypoglycemic' (that is, it had a hypoglycemic event) if any glucose during the day was < 70 mg/dl, and 'hyperglycemic' if any glucose during the day was > 299 mg/dl.

The glucometrics are:

• the mean {of all the patient-day means}
• the proportion of patient-day means within the goal range
• the proportion of patient-days tagged hypoglycemic
• the proportion of patient-days tagged hyperglycemic.

The figure below illustrates metrics for a cohort of three patients.

Note that there are a total of thirteen patient-days: five each for patients 1 and 2 and three for patient 3.
However, glucose was monitored on only 11 of those days. So for metrics calculations, we use only those eleven patient-days.

 

Here's a sample glucometrics report with explanation.