Key findings from the validation of algorithms for diabetes case definition 6
Key findings from case definitions for diabetes type ascertainment 7
Conclusion 7
1. Introduction 9
Aim of the project 10
Structure of the report 10
2. Data source and methods 11
MedicineInsight 11
Conditions 12
Study population 12
Study period 12
Markers of diabetes status 15
Prescriptions 15
HbA1c tests from pathology results 16
Medicare Benefits Schedule service items 16
Measures of validity and reference standard 17
Reference standard 17
Performance characteristics 18
Limitations 20
3. Diabetes case identification using markers of diabetes status 22
Diabetes case definitions 22
Methods 23
Findings from validation of the algorithms for diabetes case definition 25
Single diabetes marker approach 25
Combined data approaches for diabetes case identification 29
Potential algorithms for diabetes case definition 33
4. Diabetes type ascertainment 35
Type 1 and type 2 diabetes case definitions 35
Methods 36
Findings from validation of the algorithms for type 2 diabetes case definition 37
Single diabetes marker approach 37
Combined data approaches 40
Potential algorithms for type 2 diabetes case definition 43
Findings from validation of the algorithms for type 1 diabetes case definition 45
Single diabetes marker approach 45
Combined data approaches 46
Potential algorithms for type 1 diabetes case definition 48
5. Recording of diabetes markers before and after diagnosis of diabetes 50
Methods 50
Findings 50
People with diabetes markers recorded pre- and post-diagnosis 51
Pre- and post-diagnosis records for diabetes markers 53
Conclusion 55
Recommendations 56
Appendix A: Definitions 57
Appendix B: Sociodemographic characteristics of study populations 59
Acknowledgements 62
Abbreviations 63
Symbols 65
Glossary 66
References 68
Tables
Table 3.1. Misclassification risk and predictive power of diabetes status based on number of prescriptions for individual diabetes medicine class 26
Table 3.2. Predictive power of diabetes status based on number of prescriptions for any diabetes medicine 27
Table 3.3. Predictive power of diabetes status based on number of prescriptions for any diabetes medicines (excluding people with only metformin prescriptions) 27
Table 3.4. Predictive power of diabetes status based on number of HbA1c tests recorded 28
Table 3.5. Predictive power of diabetes status based on number of HbA1c tests each with 6 months gap of another HbA1c test 28
Table 3.6. Misclassification risk and predictive power of diabetes status based on number of diabetes-related MBS items (excluding MBS items for HbA1c) for... 29
Table 3.7. Predictive power of diabetes status based on number of diabetes-related MBS items (excluding MBS items for HbA1c) 29
Table 3.8. Sensitivity and PPV based on number of prescriptions for any diabetes medicine and HbA1c tests 30
Table 3.9. Predictive power of diabetes status based on combination of prescriptions for any diabetes medicine or HbA1c tests 31
Table 3.10. Predictive power of diabetes status based on combination of diabetes-related prescriptions or MBS items (excluding MBS items for HbA1c) 32
Table 3.11. Predictive power of diabetes status based on number of prescriptions for diabetes medicines or HbA1c tests or diabetes-related MBS items (excluding... 33
Table 3.12. Validation of diabetes case definition algorithms against people identified with diabetes using MedicineInsight standard definition; all study population... 34
Table 4.1. Misclassification risk and predictive power of type 2 diabetes status based on number of prescriptions for individual diabetes medicines 38
Table 4.2. Predictive power of type 2 diabetes status based on number of prescriptions for any diabetes medicines 38
Table 4.3. Predictive power of type 2 diabetes status based on number of prescriptions for any diabetes medicines (excluding insulin) 39
Table 4.4. Predictive power of type 2 diabetes status based on number of HbA1c tests 39
Table 4.5. Predictive power of type 2 diabetes status based on number of diabetes-related MBS items (excluding MBS items specific for HbA1c tests) 40
Table 4.6. Sensitivity and PPV based on number of prescriptions for diabetes medicine (excluding people with only insulin prescriptions) and HbA1c tests 41
Table 4.7. Predictive power of type 2 diabetes status based on number of prescriptions for diabetes medicines or HbA1c tests 42
Table 4.8. Predictive power of type 2 diabetes status based on number of prescriptions for diabetes medicines or diabetes-related MBS items (excluding MBS... 42
Table 4.9. Predictive power of type 2 diabetes status based on number of prescriptions for diabetes medicines or diabetes-related MBS items or HbA1c tests 43
Table 4.10. Validation of type 2 diabetes definition algorithms against people identified with type 2 diabetes using MedicineInsight standard definition; all study... 44
Table 4.11. Predictive power of type 1 diabetes status based on number of prescriptions for individual diabetes medicines 45
Table 4.12. Sensitivity and PPV based on number of prescriptions for insulin (with/without other diabetes medicines) and HbA1c tests 47
Table 4.13. Predictive power of type 1 diabetes status based on number of insulin prescriptions or diabetes-related MBS items (excluding MBS items for HbA1c) 47
Table 4.14. Predictive power of type 1 diabetes status based on number of prescriptions for insulin only (without other diabetes medicines) or diabetes-related... 48
Table 4.15. Validation of type 1 diabetes definition algorithms against people identified with type 1 diabetes using MedicineInsight standard definition; study population... 49
Table 5.1. Proportion of the diabetes incident cohort who had a minimum of one diabetes marker and proportion of those with at least one HbA1c test, diabetes-related... 52
Table 5.2. Proportion of all records for each diabetes marker for the diabetes incident cohort recorded pre- or post-diagnosis period 53
Figures
Figure 2.1. Study period for the cohorts and the markers of diabetes status in MedicineInsight 14
Figure 2.2. Number of diabetes-related prescriptions (excluding repeats) 16
Figure 2.3. Number of diabetes-related Medicare services recorded 17
Figure 2.4. Reference standard/standard definition for diabetes, type 1 and type 2 diabetes 18
Boxes
Box 2.1. Calculating measures of validity 19
Box 3.1. Examples of published methods in literature which have been used to identify diabetes cases using electronic medical records or administrative data sources 23
Box 3.2. Assessment approach and records thresholds 24
Box 3.3. Approaches used to identify diabetes case definitions 25
Box 4.1. Examples of published methods in literature which have been used to identify type 1 or type 2 diabetes cases using electronic medical records or administrative... 36
Appendix Tables
Table A1. Diabetes definition 57
Table A2. ATC codes used to identify diabetes medicines 57
Table A3. Diabetes-related Medicare Benefits Schedule item numbers included in the study 58
Table B1. Sociodemographic characteristics of study populations 60