Background (Dis)agreements between self-reports of prior clinical diagnoses and objective assessments of disease status provide potentially important insights into three specific processes: (mis)communication of clinical diagnoses; (un)awareness of disease risk; and (non)modification of disease risk post-diagnosis. This study explored these processes by examining whether sociodemographic, anthropometric and lifestyle-related factors are associated with self-reports of diabetes diagnoses that (dis)agree with subsequent objective assessments.

Methods (Dis)agreements between self-reported clinical diagnoses of diabetes and objective assessments thereof (elevated HbA1c values and/or anti-diabetic medication), based on questionnaire and ‘nurse visit’ data collected for n = 13,258 participants during Waves 1–3 of the UK Household Longitudinal Study, were used to classify each participant’s self-reported diabetes status as ‘true positive’ (TP), ‘true negative’ (TN), ‘false positive’ (FP) or ‘false negative’ (FN). Complete data on 15 potential predictors of (dis)agreement (sex; age; educational attainment; marital status; parenthood; employment; marital status; smoking; alcohol consumption; fruit/vegetable consumption; sport-related activity; body mass index [BMI]; waist circumference; and physical and mental health) were available for n = 7770 of these participants. A directed acyclic graph, in which preceding predictors were assumed to act as confounders in any relationships between each predictor and (dis)agreement, informed multinomial logistic regression analyses in STATA SE13.1 to identify which predictors were significantly associated with an increased risk of TP, FP or FN (with TN as referent) after adjustment for confounding.

Results With the exception of smoking, fruit/vegetable consumption and waist circumference, all of the remaining predictors displayed a significant relationship with either TP, FP and/or FN after adjustment for confounding. Of these, increasing age (RRR_TP = 2.45 [95% CI: 2.16–2.77]; RRR_FP = 1.58 [95% CI: 1.33–1.88]; RRR_FN = 1.80 [95% CI: 1.45–2.23]), decreasing educational attainment (RRR_TP = 1.12 [95% CI: 1.08–1.15]; RRR_FP = 1.10 [95% CI: 1.04–1.18]; RRR_FN = 1.15 [95% CI: 1.07–1.23]), and being overweight (RRR_TP = 1.77 [95% CI: 1.24–2.54]; RRR_FP = 2.91 [95% CI: 1.40–6.05]; RRR_FN = 7.06 [95% CI: 2.16–23.08]) or obese (RRR_TP = 4.60 [95% CI: 3.27–6.48]; RRR_FP = 5.78 [95% CI: 2.83–11.79]; RRR_FN = 12.75 [95% CI: 3.92–41.45]) were all associated with an increased risk of TP, FP and FN; while the only unique predictors of each category of (dis)agreement were: marital status and employment status for TP (risk being elevated for divorced/separated/widowed and unemployed participants, respectively); and parenthood for FP (risk being elevated for participants with one or more child).

Discussion Many of the same sociodemographic, anthropometric and lifestyle factors (particularly: age, educational attainment and BMI) associated with objective assessments of diabetes (be that TP or FN) are also associated with self-reported clinical diagnoses of individuals who (no longer) display haematological or medication-related evidence of diabetes (i.e. FP). However, there may also be specific markers for distinguishing between individuals who are un/aware they currently have diabetes and those with a past diagnosis but no current symptoms/medication.