Description of a New Predictive Modeling Approach That Correlates the Risk and Associated Cost of Well-Defined Diabetes-Related Complications With Changes in Glycated Hemoglobin (HbA1c)

Abstract

The modeling approach described here is designed to support the development of spreadsheet-based simple predictive models. It is based on 3 pillars: association of the complications with HbA1c changes, incidence of the complications, and average cost per event of the complication. For each pillar, the goal of the analysis was (1) to find results for a large diversity of populations with a focus on countries/regions, diabetes type, age, diabetes duration, baseline HbA1c value, and gender; (2) to assess the range of incidences and associations previously reported. Unlike simple predictive models, which mostly are based on only 1 source of information for each of the pillars, we conducted a comprehensive, systematic literature review. Each source found was thoroughly reviewed and only sources meeting quality expectations were considered. The approach allows avoidance of unintended use of extreme data. The user can utilize (1) one of the found sources, (2) the found range as validation for the found figures, or (3) the average of all found publications for an expedited estimate. The modeling approach is intended for use in average insulin-treated diabetes populations in which the baseline HbA1c values are within an average range (6.5% to 11.5%); it is not intended for use in individuals or unique diabetes populations (eg, gestational diabetes). Because the modeling approach only considers diabetes-related complications that are positively associated with HbA1c decreases, the costs of negatively associated complications (eg, severe hypoglycemic events) must be calculated separately.

Keywords: HbA1c; approach; complications; cost; diabetes; model.

Figure 1.

Flow diagram of reviewed studies. (A) The literature search in bibliographic databases retrieved a total of 4291 citations for all 3 research questions. After electronic removal of 1064 duplicates, 3227 abstracts remained for screening against inclusion/exclusion criteria. (B) The search for sources describing the relationship of HbA1c level and diabetes-related complications (search 1) provided a total of 1394 citations for the screening against inclusion/exclusion criteria. Out of these 1183 abstracts were excluded and 211 citations were found eligible for the screening on full-text level. Of these full-text publications, 97 articles were excluded and 114 were found eligible. The majority of the eligible publications provided relationship results concerning the most common complications. Relationship results concerning nephropathy, retinopathy, cardiovascular diseases, hypoglycemia, and neuropathy were found in 47, 22, 15, 14, and 10 publications. The other complications were found in relatively few publications. (C) The search for sources with regard to incidence rates of diabetic complications (search 2) provided a total of 1391 citations for the screening against inclusion/exclusion criteria. Out of these, 1130 abstracts were excluded and 261 citations were reviewed on the full-text level. In total, 190 full-text articles were considered as useful for this research questions and 71 articles were excluded. The majority of the eligible publications provided results with regard to incidences of the most common complications: hypoglycemia (n = 105), nephropathy (n = 64), cardiovascular diseases (n = 61), and retinopathy (n = 49). (D) The search for sources focusing on diabetes-related complication costs (search 3) resulted in 786 citations for the screening against inclusion/exclusion criteria. Of these, 582 abstracts were excluded and 204 citations were reviewed on full-text level. Seventy-eight articles were excluded and 126 were found eligible. The majority of eligible publications provided cost results concerning all diabetes-related complications (n = 43) or several complications bundled together (n = 24). The cost results concerning nephropathy revealed 18 publications while 16 publications have cost results attributed to hypoglycemia. Hospitalization costs for diabetes management were found in 18 publications. The other complications were found in relatively few publications.

Figure 2.

Three different deviations found for the same data (DCCT patients) by 3 analysts.^-

Figure 3.

Sensitivity analyses for number and cost of prevented events.