Biology Reference
In-Depth Information
In our study, we determined the LBGI for 600 people with type 1
(N
323) diabetes (the same data used above).
Each person had collected three to five measurements per day for 4 to
6 weeks. As in the study described in the previous section, patients
also recorded the date and time of all SH episodes during the following
6 months.
¼
277) and type 2 (N
¼
1. Developing and Testing a More Sophisticated Model
for Predicting Severe Hypoglycemia
In developing the model, we first divided the 600 LBGI values into 15
groups of 40 subjects each, with consecutive risk ranges. The resulting
intervals had the following cutoffs: 0.1, 0.5, 0.8, 1.1, 1.4, 1.7, 2.1, 2.5, 2.9,
3.5, 4.2, 5.0, 6.0, and 7.8. Thus, the lowest-risk range is LBGI < 0.1, and
the highest-risk range is LBGI
>
7.8. We numbered the classes 0, 1, 2,
,
...
14. For example, class 1 corresponds to the range 0.1
LBGI < 0.5;
class 2 corresponds to the range 0.5
LBGI < 0.8; and so on.
Next, we determined the probability that a person from each class had
an episode of SH in the following month. Suppose 14 of the 40 subjects in
class 5 reported at least one episode of SH. Then 14/40
0.35 of the
people from class 5 had an episode or, said another way, the (empirical)
probability of a person in class 5 having at least one episode of SH is
0.35. With this, we have the point (5,0.35); the first coordinate is the class,
and the second coordinate is the probability for at least one episode of
SH in that class. We do this for each of the 15 classes and obtain the
points shown by black triangles in Figure 5-13.
¼
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
0.00
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4
Risk level
P(>=1SH 1month)
P(>=1MH 1month)
Emp MH
Emp SH
FIGURE 5-13.
Probability for at least one SH episode within 1 month.
