Skip to contents

Algorithm

Source: vignettes/algorithm.Rmd
algorithm.Rmd

General description

A more complete description of the classifier is found in Anders Aasted Isaksen’s PhD Thesis as well as the validation paper (1). The description of the algorithm below is a brief and concise version of those documents.

The overall output of this algorithm is first to classify those with diabetes, then to further classify and check if the individuals might have type 1 diabetes, otherwise classify as type 2 diabetes.

The initial diabetes classification is defined as the second occurrence of any of the listed inclusion events. Wherever possible, all available data for each event is used, except for the purchases of glucose-lowering drugs, since the data on obstetric diagnoses necessary to censor glucose-lowering drug purchases is only complete from 1997 onwards. Inclusion criteria are:

  1. HbA1c measurements of ≥48 mmol/mol.
  2. Hospital diagnoses of diabetes.
  3. Diabetes-specific services received at podiatrist.
  4. Purchase of glucose-lowering drugs.

Exclusions are:

  1. HbA1c samples:
    • Samples taken during pregnancies, as that could be a potential gestational diabetes mellitus.
  2. Glucose-lowering drugs:
    • Brand drugs for weight loss, e.g. Saxenda.
    • Purchases during pregnancies, as that is a potential treatment for gestational diabetes mellitus.
    • Metformin for women below age 40, as that could be a treatment for polycystic ovary syndrome.

Classifying type 1 diabetes

Diabetes type is classified as either T1D or T2D based on patterns of purchases of insulin drugs (including analogues) and hospital primary diagnoses of T1D and T2D.

Classification as T1D requires an individual to fulfill either of the following criteria:

  1. Must have purchased only insulin drugs and never any other type of glucose-lowering drugs, and have at least one diagnosis of T1D
  2. Must have a majority of T1D diagnoses from endocrinological departments (or from other medical departments, in the absence of contacts to endocrinological departments), and a purchase of insulin within 180 days after onset of diabetes, with insulin contributing at least two thirds of all defined daily doses of glucose-lowering drugs purchased.

In populations generated on a fixed index date (such as the cross-sectional studies associated with this dissertation), individuals classified as T1D cases must have purchased insulin drugs in the last year prior to the index date.

Classifying type 2 diabetes

Individuals not classified as T1D cases are classified as T2D cases.

Detailed and technical description

Below are the technical, exact implementation of the above description. These are the logical conditions and exact variables and registers used that results in the classification of diabetes status and types. They are shown in the form of tables for each register with a table at the end showing the logic that is applied across registers.

lpr_diag

name title logic comments
lpr2_is_needed_code LPR2 codes used throughout the algorithm c_diag =~ ‘^(DO0[0-6]|DO8[0-4]|DZ3[37]|DE1[0-4]|249|250)’ AND (c_diagtype == ‘A’ OR c_diagtype == ‘B’) A c_diagtype of'A' means primary diagnosis, while ‘B’ means secondary diagnosis.
lpr2_is_diabetes_code LPR2 diagnoses codes for any diabetes c_diag =~ ‘^(DE1[0-4]|249|250)’
lpr2_is_t1d_code LPR2 diagnoses codes for T1D c_diag =~ ‘^(DE10|249)’
lpr2_is_t2d_code LPR2 diagnoses codes for T2D c_diag =~ ‘^(DE11|250)’
lpr2_is_pregnancy_code LPR2 diagnoses codes for pregnancy-related outcomes c_diag =~ ‘^(DO0[0-6]|DO8[0-4]|DZ3[37])’ These are recorded pregnancy endings like live births and miscarriages.
lpr2_is_primary_dx LPR2 primary diagnosis c_diagtype == ‘A’

lpr_adm

name title logic comments
lpr2_is_endocrinology_dept LPR2 endocrinology department c_spec == 8 TRUE when the department where the recorded diagnosis was endocrinology.
lpr2_is_medical_dept LPR2 other medical department c_spec %in% c(1:7, 9:30) TRUE when the diagnosis was recorded at a medical department other than endocrinology.

diagnoser

name title logic comments
lpr3_is_needed_code LPR3 codes used throughout the algorithm diagnosekode =~ ‘^(DO0[0-6]|DO8[0-4]|DZ3[37]|DE1[0-4])’ AND (diagnosetype == ‘A’ OR diagnosetype == ‘B’) AND (senere_afkraeftet == ‘Nej’) A diagnosekode means primary diagnosis and senere_afkraeftet means diagnosis was later retracted.
lpr3_is_primary_dx LPR3 primary diagnosis diagnosetype == ‘A’
lpr3_is_t1d_code LPR3 diagnoses codes for T1D diagnosekode =~ ‘^(DE10)’
lpr3_is_t2d_code LPR3 diagnoses codes for T2D diagnosekode =~ ‘^(DE11)’
lpr3_is_diabetes_code LPR3 diagnoses codes for diabetes diagnosekode =~ ‘^DE1[0-4]’ This is a general diabetes code, not specific to T1D or T2D.
lpr3_is_pregnancy_code ICD-10 diagnoses codes for pregnancy-related outcomes diagnosekode =~ ‘^(DO0[0-6]|DO8[0-4]|DZ3[37])’ These are recorded pregnancy endings like live births and miscarriages.

kontakter

name title logic comments
lpr3_is_endocrinology_dept LPR3 endocrinology department hovedspeciale_ans == ‘medicinsk endokrinologi’ TRUE when the department is endocrinology.
lpr3_is_medical_dept LPR3 medical department hovedspeciale_ans %in% c(‘blandet medicin og kirurgi’, ‘intern medicin’, ‘geriatri’, ‘hepatologi’, ‘hæmatologi’, ‘infektionsmedicin’, ‘kardiologi’, ‘medicinsk allergologi’, ‘medicinsk gastroenterologi’, ‘medicinsk lungesygdomme’, ‘nefrologi’, ‘reumatologi’, ‘palliativ medicin’, ‘akut medicin’, ‘dermato-venerologi’, ‘neurologi’, ‘onkologi’, ‘fysiurgi’, ‘tropemedicin’) TRUE when the department is other medical departments (than endocrinology).

lab_forsker

name title logic comments
is_hba1c_over_threshold HbA1c values over threshold (analysiscode == ‘NPU27300’ AND value >= 48) OR (analysiscode == ‘NPU03835’ AND value >= 6.5) Is the IFCC units for NPU27300 and DCCT units for NPU03835.

lmdb

name title logic comments
is_gld_code ATC codes for glucose-lowering drugs (GLDs) atc =~ ‘^A10’ AND NOT (atc =~ ‘^(A10BJ|A10BK01|A10BK03)’) GLP-RAs or dapagliflozin/empagliflozin drugs are not kept.

Across register logic

name title logic comments
is_not_within_pregnancy_interval Events that are not within a potential pregnancy interval NOT (has_pregnancy_event AND date >= (pregnancy_event_date - weeks(40)) AND date <= (pregnancy_event_date + weeks(12))) OR is.na(pregnancy_event_date) The potential pregnancy interval is defined as 40 weeks before and 12 weeks after the pregnancy event date.
is_podiatrist_services Podiatrist services speciale =~ ‘^54’ AND barnmak == 0 When barnmak == 0, the PNR belongs to the recipient of the service. When barnmak == 1, the PNR belongs to the child of the individual.
is_not_metformin_for_pcos Metformin purchases that aren’t potentially for the treatment of PCOS NOT (koen == 2 AND atc =~ ‘^A10BA02$’ AND ((date - foed_dato) < years(40) OR indication_code %in% c(‘0000092’, ‘0000276’, ‘0000781’))) Woman is defined as 2 in koen.

References

1.
Isaksen AA, Sandbæk A, Bjerg L. Validation of register-based diabetes classifiers in danish data. Clinical Epidemiology. 2023 May;Volume 15:569–81.