Measured Blooc

Figure 7.1. Error Grid analysis (EGA). Blood glucose measured in mg/dl average deviation between the two can be low, even when patients make errors with potentially serious clinical implications. To address the question of the impact of BG detection on diabetes regulation, we developed a classification procedure, the Error Grid Analysis34, which quantifies the clinical significance of patient accuracy.* Figure 7.1 shows the Error Grid Analysis (EGA), which involves plotting patients' BG estimates (y axis) in comparison to measured BG (x axis), so that each estimate falls into one of 10 categories or zones. Each zone represents the potential clinical outcome of taking action to adjust BG based on the subjective estimate.

The Error Grid is divided by a regression line that represents perfect agreement between estimated and actual BG. Data points falling above and below this line represent overestimates and underestimates of actual BG, respectively. The upper and lower A zones include estimates considered to be clinically acceptable; that is, the estimate is either within 20% of actual BG or both estimated and actual BG are below 3.9 mM (70 mg/dl). Estimates falling into the upper and lower B zones deviate more than 20% from actual BG, but are considered to be benign errors, since it is unlikely that patients would take action that was clinically dangerous based on these estimates, if action were taken, the clinical implications would be minimal. The C, D and E zones are considered to be clinically significant errors. C zone errors represent potential unnecessary corrections of BG because it estimated to be information concerning the purchase of computer software for Error Grid Analysis is available from the author.

too high (upper C zone) or too low (lower C zone) when actual glucose is in an acceptable range. Taking action to treat BG based on these estimates could result in hypo- or hyperglycaemia. D zone errors occur when estimated BG is in an acceptable range but actual BG is too low (upper D zone) or high (lower D zone). These failures to detect hypo- or hyperglycaemia can lead to a failure to provide needed self-treatment. Finally, E zone estimates can lead to erroneous self-treatment, such as taking action to lower BG when hypogly-caemic (upper E zone) or to raise BG when hyperglycaemic (lower E zone).

Our first study of BG detection assessed patient accuracy under two conditions*, during hospital testing (while BG levels were manipulated via insulin and glucose infusion) and at home (while following normal daily routines)34. During hospital testing and BG manipulation, external cues such as time of day were irrelevant, forcing patients to rely on symptom feedback to estimate BG. In contrast, during home testing patients had access to external cues. As Figure 7.2 shows, patients made significantly more accurate estimates during home testing (A zones = 61%) as compared to hospital testing (A zones = 46%). Although the majority of estimates were clinically acceptable or benign errors, patients also made a substantial number of clinically significant errors. A total of 19% and 13% of estimates fell in the C, D and E zones in the hospital and home conditions, respectively. By far the most common error was failure to detect extreme BG levels (D zones), with C and E zone errors being rare. Unawareness of hypoglycaemia occurred more frequently than unawareness of hyperglycaemia, with patients detecting BGs < 3.9 mM only about 50% of the time. However, in a more recent laboratory study25, BG levels were raised to a more extreme hyperglycaemic range (21.1 mM). These researchers found more errors in detecting high BG as compared to low BG. While only 17% of patients made clinically significant errors in detecting hypoglycaemia, 66% made these errors in their detection of hyperglycaemia—errors that included unawareness that BG levels were high (lower D zone) and believing BG was hypoglycaemic when it was quite high (lower E zone).

Other studies, by our research group and others, have used the EGA to assess BG detection, often employing a summary measure of the EGA called the accuracy index (AI). The AI is computed by subtracting the percentage of clinically significant errors (C, D and E zones) from the percentage of A zone estimates. In studies of adults with type 1 diabetes and a history of using SMBG, AI scores have ranged from 35% to 60%37-39. Accuracy appears to be

*The basic method for assessing accuracy of detection involves asking the patient, who is blind to actual glucose level, to estimate current BG on the basis of available cues, followed by an objective measure of glucose. BG estimates and measures can be recorded on paper forms or entered directly into computers. In the hospital assessment, BG estimates and measures were obtained every 10 minutes over a period of several hours and entered into a hand-held computer. In the home assessment, patients recorded their estimates on paper forms before SMBG several times each day over a period of 1-2 weeks.

Figure 7.2. Percentage of patient blood glucose (BG) estimates falling into each EGA zone

significantly poorer in younger age groups. For example, the first study of adolescents14 found that 55% of BG estimates were clinically accurate; however, since that study did not compute clinically significant errors, AI scores could not be computed. A subsequent study40 found poorer accuracy in both young adults (AI = 32%) and adolescents (AI = 7%). Children with type 1 diabetes and their parents exhibit even poorer accuracy, with average AI scores of — 1.05% and 5.0%, respectively41. This means that children and parents make clinically serious errors almost as often as clinically accurate estimates. Children also differed from adult patients in the types of errors they made. As Figure 7.3 shows, children made a much higher rate of lower E zone errors than adult patients. Thus, children often believed they were hypoglycaemic when in fact their BG was high, which could lead them to take action to raise BG when it is already too high. Several factors likely contribute to children's poor ability to detect BG extremes and the tendency to make more lower E zone errors; for example, developmental differences in the cognitive skills involved in discriminating physical symptoms. In addition, parental concern about and focus on immediate treatment may increase the salience of hypoglycaemia, and there are important reinforcement contingencies, since children typically are given some sort of sweet food or drink to treat low BG.

Patients who report reduced hypoglycaemic awareness also show poorer accuracy30. They detect only about 33% of BG readings below 3.9 mM, compared to adults without reduced awareness who detect, on average, 50% of their low BGs. Overall accuracy (AI) is also significantly lower (15%), closer to that seen in younger patient groups.

0 0

Post a comment