Per Winterdijk¹, Pim Dekker¹, Christine Fransman¹, Henk-Jan Aanstoot¹,

Giesje Nefs^1,2

¹ Diabeter, ² Radboudumc, Diabeter Centrum Amsterdam, Tilburg University

Background: Diabeter is implementing yearly evaluations of clinical and mental health. These data are used in a shared decision-making discussion to improve persons’ diabetes care. With respect to Person Reported Outcome (PRO) measurement, we identified 12 psychosocial domains that affect diabetes care and can be improved. For use in everyday diabetes care a brief and comprehensive PROM set is required.

Aims: To measure multiple domains of PROs in type 1 diabetes (T1D) care while minimizing completion burden. We examined (a) which index questions best predict if more questions are needed for the domains depression, anxiety and diabetes-specific worries, based on elevated scores on existing in-depth questionnaires, and (b) the most informative in-depth items.

Methods: Participants ≥16 years with T1D completed a survey including five candidate index questions per domain and in-depth questionnaires: depressive symptoms PHQ-9, WHO-5; anxiety GAD-7; diabetes-specific worries PAID-20, DSPAV, HFS-W, HBS-W. Per domain, we selected the index question best predicting elevated scores on the in-depth questionnaires using the highest negative predictive value (NPV). Item response theory (partial credit; maximal information over theta range guiding item selection) identified the most informative in-depth questions. With these we repeated the index question selection.

Results: 456 people completed the survey (63% women; mean±SD age: 34±14 years; diabetes duration 18±11 years; HbA_1c 7.1±1.1%). Index questions, with highest NPV, were ‘Are you unhappy, dissatisfied or have less energy to do anything?’(depressive symptoms; NPV=0.49), ‘Have you been feeling anxious, tense or worried lately’, (anxiety; NPV=0.90), and ‘Do you often feel frustrated or find it difficult when thinking about hypos, hypers, your future, injecting or finger pricking?’ (diabetes-specific worries; NPV=0.23). Most informative were 9 items for depressive symptoms (originally 14), 4 for anxiety (7) and 28 (64) for diabetes-specific worries. For the IRT reduced in-depth set, the optimal index questions were the same (with NPV 0.52, 0.94, 0.36).

Conclusion: Index questions predicting elevated scores on IRT-shortened questionnaires were the same as for the original in-depth questionnaires, with slightly better NPV. Using index questions followed by short in-depth question sets if required offers an opportunity to minimize questionnaire burden in clinical practice.