Systems and methods for pattern recognition in diabetes management
A diabetes management system or process is provided herein that may be used to analyze and recognize patterns for a large number of blood glucose concentration measurements and other physiological parameters related to the glycemia of a patient. In particular, a method of monitoring glycemia in a patient may include storing a patient's data on a suitable device, such as, for example, a blood glucose meter. The patient's data may include blood glucose concentration measurements. The diabetes management system or process may be installed on, but is not limited to, a personal computer, an insulin pen, an insulin pump, or a glucose meter. The diabetes management system or process may identify a plurality of pattern types from the data including a testing/dosing pattern, a hypoglycemic pattern, a hyperglycemic pattern, a blood glucose variability pattern, and a comparative pattern. After identifying a particular pattern with the data management system or process, a warning message may be displayed on a screen of a personal computer or a glucose meter. Other messages can also be provided to ensure compliance of any prescribed diabetes regiments or to guide the patient in managing the patient's diabetes.
1. A method of monitoring glycemia in a patient with a glucose meter that includes at least a power source, microprocessor, memory and display, the method comprising:
measuring, with a microprocessor of the meter, a plurality glucose concentrations of a patient with the glucose meter to provide a plurality of glucose measurements;
storing in the memory of the meter, the plurality of glucose measurements of the patient;
using the microprocessor, generating, from the memory, statistically significant patterns from the patient's glucose measurements, the patterns indicative of hypoglycemia, hyperglycemia, or excessive glucose variability by time of day, by day in a week, both by time of day and day of week, or at different time intervals which are selected from a group consisting of a time interval between visits to a physician, a time interval between visits to a clinician, a time interval between different prescribed therapies, and combinations thereof, the generating comprising:
determining a hypoglycemic pattern by:
obtaining a number of glucose measurements over a total time period;
dividing the total time period into a plurality of time intervals;
determining a percentage of hypoglycemic incidence for each of the time intervals which recurs daily and is equal to about one eighth of a day;
determining whether the percentage of hypoglycemic incidence for at least one of the time intervals is statistically significantly different;
displaying a message upon one of the patterns being indicative of a pattern of glycemia outside at least a predetermined range for such pattern with the display;
using the microprocessor, utilizing a chi-squared test to determine if any of the time intervals is statistically significantly different wherein the chi-squared test uses a confidence level ranging from about 95% to about 99%, the number of glucose measurements is greater than about 27, the chi-squared test is of the form:
χ
2
=
∑
i
=
1
n
(
L
i
-
L
i
,
pre
)
2
L
i
.
pre
+
∑
i
=
1
n
(
L
i
′
-
L
i
,
pre
′
)
2
L
i
.
pre
′
where χ 2 =chi-squared, i represents a particular time interval, n is a total number of time intervals, L i is a number of substantially hypoglycemic glucose concentration measurements that occur during time interval i, L i,pre is a predicted number of substantially hypoglycemic glucose concentration measurements that will occur during time interval i, and L′ i,pre is a predicted number of non-hypoglycemic glucose concentration measurements that will occur during interval time i, L i,pre using an estimation equation, the estimation equation comprising:
L
i
,
pre
=
∑
i
=
1
n
L
i
∑
i
=
1
n
N
i
*
N
i
where N i represents the total number of glucose concentration measurements performed during timer interval i;
comparing a calculated χ 2 to a χ 2 value in a table based on a number of degrees of freedom for each of said time intervals i, wherein the table includes a plurality of conditions that are related to at least two outcomes, one of which is a hypoglycemic outcome and the other of a non-hypoglycemic outcome; and
determining that at least one of the time intervals are statistically significantly different if the calculated χ 2 is greater than the χ 2 value on the table.
2. The method of claim 1 , further comprising:
calculating Z i using a Z test, the Z test comprising:
Z
i
=
(
L
i
-
L
i
,
pre
)
SE
i
where Z i represents a Z value at a particular time interval i and SE i represents a standard error for a particular time interval i, the standard error SE i comprising:
SE
i
=
1
N
i
*
L
i
,
pre
*
(
N
i
-
L
i
,
pre
)
comparing a calculated Z i to a Z value in the table; and
identifying that one of the time intervals are statistically significantly different if the calculated Z i is greater than the Z value of about two.
3. The method of claim 1 , wherein the generating comprises determining hyperglycemic patterns by:
obtaining a number of glucose measurements over a total time period;
dividing the total time period into a plurality of time intervals;
determining percentage of hyperglycemic incidence for each of the time intervals which recurs daily and is equal to about one eighth of a day;
determining whether the percentage of hyperglycemic incidence for at least one of the time intervals is statistically significantly different with a chi squared test χ 2 with a confidence level ranging from about 95% to about 99% that comprises:
χ
2
=
∑
i
=
1
n
(
H
i
-
H
i
,
pre
)
2
H
i
,
pre
+
∑
i
=
1
n
(
H
i
′
-
H
i
,
pre
′
)
2
H
i
,
pre
′
where χ 2 =chi-squared, i represents a particular time interval, n is a total number of time intervals, H i is a number of substantially hyperglycemic glucose concentration measurements that occur during time interval i, H i,pre is a predicted number of substantially hyperglycemic glucose concentration measurements that will occur during time interval i, where H i,pre comprises:
H
i
,
pre
=
∑
i
=
1
n
H
i
∑
i
=
1
n
N
i
*
N
i
where N i represents the total number of glucose concentration measurements performed during timer interval i, and H′ i,pre is a predicted number of non-hyperglycemic glucose concentration measurements that will occur during time interval i;
comparing a calculated χ 2 to a χ 2 value in a table based on a number of degrees of freedom for each of said time intervals i; and
determining that at least one of the time intervals are statistically significantly different if the calculated χ 2 is greater than the χ 2 value on the table;
identifying one of the time intervals as being statistically significantly different using a Z test comprising:
Z
i
=
(
H
i
-
H
i
,
pre
)
SE
i
where Z i represents a Z value at a particular time interval i and SE i represents a standard error for a particular time interval i;
comparing a calculated Z i to a Z value in a table; and
identifying that one of the time intervals are statistically significantly different if the calculated Z i is greater than the Z value of about two;
calculating SE i , using a standard error equation, the standard error equation comprising:
SE
i
=
1
N
i
*
H
i
,
pre
*
(
N
i
-
H
i
,
pre
)
;
and
displaying a message indicating a high incidence of hyperglycemia occurring on at least one of the time intervals based on the chi-squared and Z tests.