A novel method for measuring patients' adherence to insulin dosing guidelines: introducing indicators of adherence

Diabetic type 1 patients are often advised to use dose adjustment guidelines to calculate their doses of insulin. Conventional methods of measuring patients' adherence are not applicable to these cases, because insulin doses are not determined in advance. We propose a method and a number of indicators to measure patients' conformance to these insulin dosing guidelines.

We used a database of logbooks of type 1 diabetic patients who participated in a summer camp. Patients used a guideline to calculate the doses of insulin lispro and glargine four times a day, and registered their injected doses in the database. We implemented the guideline in a computer system to calculate recommended doses. We then compared injected and recommended doses by using five indicators that we designed for this purpose: absolute agreement (AA): the two doses are the same; relative agreement (RA): there is a slight difference between them; extreme disagreement (ED): the administered and recommended doses are merely opposite; Under-treatment (UT) and over-treatment (OT): the injected dose is not enough or too high, respectively. We used weighted linear regression model to study the evolution of these indicators over time.

We analyzed 1656 insulin doses injected by 28 patients during a three weeks camp. Overall indicator rates were AA = 45%, RA = 30%, ED = 2%, UT = 26% and OT = 30%. The highest rate of absolute agreement is obtained for insulin glargine (AA = 70%). One patient with alarming behavior (AA = 29%, RA = 24% and ED = 8%) was detected. The monitoring of these indicators over time revealed a crescendo curve of adherence rate which fitted well in a weighted linear model (slope = 0.85, significance = 0.002). This shows an improvement in the quality of therapeutic decision-making of patients during the camp.

Our method allowed the measurement of patients' adherence to their insulin adjustment guidelines. The indicators that we introduced were capable of providing quantitative data on the quality of patients' decision-making for the studied population as a whole, for each individual patient, for all injections, and for each time of injection separately. They can be implemented in monitoring systems to detect non-adherent patients.