The answer to this question is for classification as Type 1 or Type 2 in adults (age >20 years) known (or believed) to have diabetes. Distinguishing Type 1 from Type 2 diabetes in children (age < 20 years) would require a different approach and has a separate literature.
Here are citations and links to three publications that describe the use “administrative data” or data from electronic health records to classify adults as having Type 1 or Type 2 diabetes. The papers describe algorithms and tree-based classification schemes and the performance of the classification schemes in relation to a “gold standard.”
Lo-Ciganic W, Zgibor JC, Ruppert K, Arena VC, Stone RA. Identifying type 1 and type 2 diabetic cases using administrative data: a tree-structured model. J Diabetes Sci Technol. 2011 May 1;5(3):486-93. PubMed PMID: 21722564; PubMed Central PMCID: PMC3192615.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3192615/
Klompas M, Eggleston E, McVetta J, Lazarus R, Li L, Platt R. Automated detection and classification of type 1 versus type 2 diabetes using electronic health record data. Diabetes Care. 2013 Apr;36(4):914-21. doi: 10.2337/dc12-0964. Epub 2012 Nov 27. PubMed PMID: 23193215; PubMed Central PMCID: PMC3609529.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3609529/
Schroeder EB, Donahoo WT, Goodrich GK, Raebel MA. Validation of an algorithm for identifying type 1 diabetes in adults based on electronic health record data. Pharmacoepidemiol Drug Saf. 2018 Oct;27(10):1053-1059. doi: 10.1002/pds.4377. Epub 2018 Jan 2. PubMed PMID: 29292555; PubMed Central PMCID: PMC6028322.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6028322/
The paper by Lo-Ciganic gives a tree-based model for predicting T1DM and T2DM cases (Figure 1).
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3192615/bin/dst-05-0486-g001.jpg
The tree-based model performed well:
“Considering T1DM as the positive category, sensitivity, specificity,
PPV, and NPV of T1DM cases for the tree-structured model were 92.8%,
99.3%, 89.5%, and 99.5%, respectively (Table 3). Approximately 7.2% of T1DM cases were misclassified as T2DM, and 0.73% of T2DM cases were
misclassified as T1DM, with an overall misclassification rate of
1.1%.”
However, this model uses information on in-patient care and whether insulin or an oral hypoglycemic drug were used as an inpatient. This information is often not available in electronic health records.
The paper by Klompas describes development of an "optimized" algorithm to identify people with Type 1 diabetes from adults with a mix of Type 1 and Type 2 diabetes. Table 3 of this paper shows the many factors examined as possible factors that distinguish Type 1 from Type 2 diabetes (e.g., age, triglycerides, prescriptions for insulin) and the sensitivity and positive predictive value of each factor in identifying Type 1 and Type 2 diabetes. The Table is useful in understanding what distinguishes Type 1 and Type 2 diabetes in adults.
The authors describe an “optimized” classification scheme for Type 1 and Type 2 diabetes using electronic health record data. The optimized algorithm used only data on ICD-9 codes, prescriptions, and laboratory tests to classify patients as having Type 1 or Type 2 diabetes (Table 4). Classified as Type 1 diabetes were people with any of the following:
“a plurality of ICD-9 codes for type 1 diabetes and a prescription for
glucagon, a plurality of ICD-9 codes for type 1 diabetes and a
negative history of prescriptions for oral hypoglycemics other than
metformin, a negative plasma C-peptide, positive diabetes autoantibody
tests, or a prescription for urine acetone test strips.”
Other people with diabetes were classified as Type 2.
The performance of the Klompas algorithm was good:
“The final algorithm flagged 73 patients [as Type 1], including all 66
patients with type 1 diabetes (raw sensitivity, 100% [95% CI 96–100];
positive predictive value, 90% [82–96]). Correcting for the sampling
strategy yielded a net population-weighted sensitivity of 100%
(100–100) and a positive predictive value of 96% (91–99).”
The paper by Schroeder and colleagues describes an evaluation of the “Klompas” algorithm using an independent data source. The authors found that:
“The Klompas algorithm identified 3,286 (4.9% of 66,690) adults with
diabetes as having type 1 diabetes. Based on chart reviews, the
overall positive predictive value was 94.5%. The requirement that the
majority of diabetes diagnosis codes be type 1 identified 3,000 (4.5%)
as having type 1 diabetes, and had a positive predictive value of
96.4%. However, the algorithm criterion involving dispensing of urine acetone test strips performed poorly, with a positive predictive value
of 20.0%.”
Schroeder and co-authors ultimately recommended that:
“In settings where C-peptide and diabetes autoantibodies lab results
values are available, we recommend using the Klompas algorithm without
the urine test strips criterion to identify type 1 diabetes in adults:
- Over 50% of diabetes codes (ICD-9 250.x0, 250.x1, 250.x2, and ICD-9 250.x3; or ICD-10 E9.xx, E10.xx) were type 1 codes (ICD-9 250.x1, 250.x3, or ICD-10 E10.xx), AND no dispensing for a non-insulin antidiabetic drug (excluding metformin)
- Over 50% of diabetes codes were type 1 codes (same codes as in #1), AND a dispensing for glucagon
- Negative C-peptide result or positive diabetes autoantibodies lab test result.