Tracking the probability for imminent hypoglycemia in diabetes from self-monitoring blood glucose (SMBG) data
A method, system and related computer program product for tracking the probability of hypoglycemia from routine self-monitoring of blood glucose (SMBG) data in patients with diabetes. A specific bivariate probability distribution of low BG events based jointly on the Low BG Index (LBGI) and the Average Daily Risk Range (ADRR) is used to predict hypoglycemia probability of occurrence from inputted SMBG data. The SMBG data is retrieved from a series of SMBG data of a patient available from the patient's glucose meter and allows tracking of the probability for future hypoglycemia over a predetermined duration, e.g., a 24 or 48 hour period. The tracking includes presentation of visual and/or numerical output, as we construction of hypoglycemia risk trajectories that would enable warning messages for crossing of predefined thresholds, such as 50% likelihood for upcoming hypoglycemia below 50 mg/dl.
1. A method for maintaining the health of a diabetic patient by preventing the occurrence of a hypoglycemic event in said patient, comprising:
obtaining self monitoring blood glucose (SMBG) readings from the patient;
measuring glycemic variability of said patient and low blood glucose (BG) of said patient based on said obtained SMBG readings;
creating in a processor a bivariate distribution that maps probability for an upcoming hypoglycemic event in said patient jointly to values of said measured glycemic variability and said measured low blood glucose (BG);
optimizing in said processor the bivariate distribution to achieve prediction of a predetermined percentage of hypoglycemic events below a predetermined BG value occurring in said patient within a predetermined future time period;
tracking in said processor the optimized distribution over time using routine SMBG readings from the patient;
outputting via said processor a message to said patient when said optimized distribution indicates a certain probability for the occurrence of a hypoglycemic event in said patient within said predetermined future time period, based on SMBG data obtained from said patient; and
causing said patient to take a physical action in response to receiving said message to prevent a hypoglycemic event from occurring in said patient.
2. The method of claim 1 , wherein the function measuring glycemic variability is Average Daily Risk Range (ADRR).
3. The method of claim 2 , wherein
ADRR
=
1
M
∑
i
=
1
M
[
LR
i
+
HR
i
]
LR i =max( rl ( x 1 i ), rl ( x 2 i ), . . . , rl ( x n i )) and
HR i =max( rh ( x 1 i ), rh ( x 2 i ), . . . , rh ( x n i )) for day# i; i =1,2, . . . M;
x 1 M , x 2 M , . . . , x n M are a series of n M SMBG readings (BG) taken on Day M;
rl ( BG )= r ( BG ) if f ( BG )<0 and 0 otherwise;
rh ( BG )= r ( BG ) if f ( BG )>0 and 0 otherwise;
r ( BG )=10 f ( BG ) 2 ;
f ( BG ,α,β)=[(ln( BG )) α −β],α,β>0.
4. The method of claim 1 , wherein the function measuring low blood glucose is Low Blood Glucose Index (LBGI).
5. The method of claim 4 , wherein
LBGI
=
1
(
n
M
-
1
+
n
M
)
[
∑
s
=
1
n
M
-
1
rl
(
x
s
M
-
1
)
+
∑
t
=
1
n
M
rl
(
x
t
M
)
]
x 1 M , x 2 M , . . . , x n M are a series of n M SMBG readings (BG) taken on Day M:
rl ( BG )= r ( BG ) if f ( BG )<0 and 0 otherwise;
r ( BG )=10 f ( BG ) 2 ;
f ( BG ,α,β)=[(ln( BG )) α −β],α,β>0.
6. The method of claim 1 , wherein said optimizing comprises determining threshold values of said functions that are effective to predict a predefined minimum percentage of all occurrences of hypoglycemic events in said patient.
7. The method of claim 6 , wherein said percentage is 50%.
8. The method of claim 6 , wherein a hypoglycemic event is determined to be BG s 50 mg/dl.
9. The method of claim 6 , wherein said predetermined future time period is a succeeding 24 hour time period.
10. The method of claim 3 , wherein the function measuring low blood glucose is Low Blood Glucose Index (LBGI).
11. The method of claim 10 , wherein
LBGI
=
1
(
n
M
-
1
+
n
M
)
[
∑
s
=
1
n
M
-
1
rl
(
x
s
M
-
1
)
+
∑
t
=
1
n
M
rl
(
x
t
M
)
]
x 1 M , x 2 M , . . . x n M are a series of n M SMBG readings (BG) taken on Day M;
rl ( BG )= r ( BG ) if f ( BG )<0 and 0 otherwise;
r ( BG )=10 f ( BG ) 2 ;
f ( BG ,α,β)=[(ln( BG )) α −β],α,β>0.
12. The method of claim 11 , wherein the probability P for upcoming hypoglycemia is given by
P
(
Hypo
|
ADRR
>
x
&
LBGI
>
y
)
=
a
(
y
)
+
(
100
-
a
(
y
)
)
(
x
b
(
y
)
)
c
(
y
)
1
+
(
x
b
(
y
)
)
c
(
y
)
a
(
y
)
=
α
a
+
β
a
y
b
(
y
)
=
α
b
+
β
b
y
c
(
y
)
=
α
c
+
β
c
(
y
δ
c
)
γ
c
1
+
(
y
δ
c
)
γ
c
.
13. The method of claim 12 , wherein
α a =15.1 range: [5,20]
β a =3.13 range: [1,5]
α b =116 range: [50,150]
β b =−5.66 range: [−10,0]
α c =2.9 range: [1,5]
β c =1 range: [1,5]
δ c =2.35 range: [1,10]
γ c =3.76 range: [1,5]
14. The method of claim 13 , wherein said mapping is based on results of a training data set obtained from a population of subjects having Type 1 diabetes.
15. The method of claim 14 , wherein said mapping maps coordinate pairs of {LBGI, ADRR} to the probability for hypoglycemia, defined as BG ≤50 mg/dl, in the subsequent 24 hours.
16. The method of claim 15 , wherein LBGI is computed from SMBG readings in the preceding 48 hours.
17. The method of claim 16 , wherein ADRR is computed from SMBG readings in the preceding 14 days.
18. The method of claim 17 , wherein the presence of a flag triggers outputting of said message, and wherein
flag
=
{
1
if
ADRR
>
x
&
LBGI
>
y
0
otherwise
Where x and y are predefined values.
19. A system for maintaining the health of a diabetic patient by preventing the occurrence of a hypoglycemic event in said patient, comprising:
a processor configured to obtain self monitoring blood glucose (SMBG) readings from the patient and to measure glycemic variability of said patient and low blood glucose (BG) of said patient based on said obtained SMBG readings;
a storage medium;
a bivariate distribution stored in said storage medium, which maps probability for an upcoming hypoglycemic event in said patient jointly to values of said measured glycemic variability and said measured low blood glucose (BG), wherein said bivariate distribution allows said processor to predict a predetermined percentage of hypoglycemic events below a predetermined BG value occurring in said patient within a predetermined future time period;
said processor being configured to optimize said bivariate distribution and to track the optimized distribution over time using routine SMBG readings received from the patient;
said processor being configured to output a message to said patient when said optimized distribution indicates a certain probability for the occurrence of a hypoglycemic event in said patient within said predetermined future time period, based on SMBG data obtained from said patient; and
in response to receiving said message said patient is caused to take a physical action to prevent a hypoglycemic event from occurring in said patient.
20. The system of claim 19 , wherein the function measuring glycemic variability is Average Daily Risk Range (ADRR).
21. The system of claim 20 , wherein
ADRR
=
1
M
∑
i
=
1
M
[
LR
i
+
HR
i
]
LR i =max( rl ( x 1 i ), rl ( x 2 i ), . . . , rl ( x n i )) and
HR i =max( rh ( x 1 i ), rh ( x 2 i ), . . . , rh ( x n i )) for day# i; i =1,2, . . . M;
x 1 M , x 2 M , . . . x n M are a series of n M SMBG readings (BG) taken on Day M;
rl ( BG )= r ( BG ) if f ( BG )<0 and 0 otherwise;
rh ( BG )= r ( BG ) if f ( BG )>0 and 0 otherwise;
r ( BG )=10 f ( BG ) 2 ;
f ( BG ,α,β)=[(ln( BG )) α −β],α,β>0.
22. The system of claim 19 , wherein the function measuring low blood glucose is Low Blood Glucose Index (LBGI).
23. The system of claim 22 , wherein
LBGI
=
1
(
n
M
-
1
+
n
M
)
[
∑
s
=
1
n
M
-
1
rl
(
x
s
M
-
1
)
+
∑
t
=
1
n
M
rl
(
x
t
M
)
]
x 1 M , x 2 M , . . . , x n M are a series of n M SMBG readings (BG) taken on Day M;
rl ( BG )= r ( BG ) if f ( BG )<0 and 0 otherwise;
r ( BG )=10 f ( BG ) 2 ;
f ( BG ,α,β)=[(ln( BG )) α −β],α,β>0.
24. The system of claim 19 , wherein said optimizing comprises determining threshold values of said functions that are effective to predict a predefined minimum percentage of all occurrences of hypoglycemic events in said patient.
25. The system of claim 24 , wherein said percentage is 50%.
26. The system of claim 24 , wherein a hypoglycemic event is determined to be BG ≤50 mg/dl.
27. The system of claim 24 , wherein said predetermined future time period is a succeeding 24 hour time period.
28. The system of claim 21 , wherein the function measuring low blood glucose is Low Blood Glucose Index (LBGI).
29. The system of claim 28 , wherein
LBGI
=
1
(
n
M
-
1
+
n
M
)
[
∑
s
=
1
n
M
-
1
rl
(
x
s
M
-
1
)
+
∑
t
=
1
n
M
rl
(
x
t
M
)
]
x 1 M , x 2 M , . . . x n M are a series of n M SMBG readings (BG) taken on Day M;
rl ( BG )= r ( BG ) if f ( BG )<0 and 0 otherwise;
r ( BG )=10 f ( BG ) 2 ;
f ( BG ,α,β)=[(ln( BG )) α −β],α,β>0.
30. The system of claim 29 , wherein the probability P for upcoming hypoglycemia is given by
P
(
Hypo
|
ADRR
>
x
&
LBGI
>
y
)
=
a
(
y
)
+
(
100
-
a
(
y
)
)
(
x
b
(
y
)
)
c
(
y
)
1
+
(
x
b
(
y
)
)
c
(
y
)
a
(
y
)
=
α
a
+
β
a
y
b
(
y
)
=
α
b
+
β
b
y
c
(
y
)
=
α
c
+
β
c
(
y
δ
c
)
γ
c
1
+
(
y
δ
c
)
γ
c
.
31. The system of claim 30 , wherein
α a =15.1 range: [5,20]
β a =3.13 range: [1,5]
α b =116 range: [50,150]
β b =−5.66 range: [−10,0]
α c =2.9 range: [1,5]
β c =1 range: [1,5]
δ c =2.35 range: [1,10]
γ c =3.76 range: [1,5]
32. The system of claim 31 , wherein said mapping is based on results of a training data set obtained from a population of subjects having Type 1 diabetes.
33. The system of claim 32 , wherein said mapping maps coordinate pairs of {LBGI, ADRR} to the probability for hypoglycemia, defined as BG s 50 mg/dl, in the subsequent 24 hours.
34. The system of claim 33 , wherein LBGI is computed from SMBG readings in the preceding 48 hours.
35. The system of claim 34 , wherein ADRR is computed from SMBG readings in the preceding 14 days.
36. The system of claim 35 , wherein the presence of a flag triggers outputting of said message, and wherein
flag
=
{
1
if
ADRR
>
x
&
LBGI
>
y
0
otherwise
Where x and y are predefined values.
37. The method of claim 1 , wherein SMBG data obtained from said patient is an individual SMBG reading.
38. The method of claim 1 , wherein SMBG data obtained from said patient is all SMBG data collected from a patient in a predetermined cycle.
39. The system of claim 19 , wherein SMBG data obtained from said patient is an individual SMBG reading.
40. The system of claim 19 , wherein SMBG data obtained from said patient is all SMBG data collected from a patient in a predetermined cycle.
41. A computer program product comprising a non-transitory computer-readable storage medium containing computer-executable instructions for maintaining the health of a diabetic patient by preventing the occurrence of a hypoglycemic event in said patient, said instructions causing a computer to:
obtain self monitoring blood glucose (SMBG) readings from the patient;
measure glycemic variability of said patient and low blood glucose (BG) of said patient based on said obtained SMBG readings;
create a bivariate distribution in said storage medium, which maps probability for an upcoming hypoglycemic event in said patient jointly to values of said measured glycemic variability and said measured low blood glucose (BG), wherein said bivariate distribution allows prediction of a predetermined percentage of hypoglycemic events below a predetermined BG value occurring within a predetermined future time period;
optimize said bivariate distribution and track the optimized distribution over time using routine SMBG readings received from the patient;
output a message to said patient when said optimized distribution indicates a certain probability for the occurrence of a hypoglycemic event in said patient within said predetermined future time period, based on SMBG data obtained from said patient; and
cause said patient in response to receiving said message to take a physical action to prevent a hypoglycemic event from occurring in said patient.
42. The computer program product of claim 41 , wherein SMBG data obtained from said patient is an individual SMBG reading.
43. The computer program product of claim 41 , wherein SMBG data obtained from said patient is all SMBG data collected from a patient in a predetermined cycle.