Table 1 Limitations, validity, and quality analysis of the data to categorize the answerability of questions

Could be answered with confidence

Number of unique individuals and total outpatient visits

n/a

Age

n/a

Sex

n/a

Types of visits (in person vs. virtual)

n/a

Laboratory markers

n/a

Answered but with documented limitations

Proportion of individuals with at least one lipid panel result recorded

Proportion of individuals with non-HDL levels recorded was lower compared to other lipid profile components, even when parts of the lipid panel are drawn and reported as a unit and not standalone tests. After consulting with the clinical biochemist, it was determined that the discrepancy is likely due to lab reporting practices. Prior to 2016, non-HDL levels were not provided by the lab and required the clinician to calculate the result manually. After 2016, there was a change in practice and the lab now reports non-HDL levels automatically. As laboratory data were analyzed between 2014–2018, non-HDL levels would not have been captured in the former half of the results, hence explaining the lower than expected proportion.

Diabetes type

Almost 1 in 4 individuals had multiple conflicting types of diabetes recorded in the EMR. These individuals are coded as “uncertain diabetes type” in our analyses as it is impossible to determine the true diagnosis without an in-depth chart review. The accuracy of coding was questioned as Diabetes in Pregnancy and Gestational Diabetes Mellitus (GDM) were both coded in males (n = 12). It is possible that some of the males documented to have GDM may have been transgender individuals.

Proportion of in-person visits with blood pressure measured

Lower than expected at a clinic where there are nurses dedicated to recording blood pressure for every in-person physician visit. This may possibly be a data entry or data capture error; we could not carry out effective data mapping to explain the discrepancy due to resource constraints.

Proportion of in-person visits with height and weight measured

Lower than expected as the dedicated nurses also measure height and weight. BMI measurements may have been missing as not all individuals had a height recorded within the study period. This may be due to height only being checked at the initial visit and not carrying forward past 365 days from previous encounters in the database to enable BMI calculation in subsequent encounters.

Length of appointments

Appointment lengths were available for clinic visits, appointments, consult letters, and chronic disease management. These were based on how long the appointments were booked for and not how long the practitioner actually spent with the individual. Some visit types, such as half-day (or longer) classes, will inflate the mean appointment lengths.

Comorbidities and complications

Identifying comorbidities and complications in both the EMR and administrative data proved to be difficult. Literature was consulted to find case definitions used by other researchers, clinicians, and organizations (e.g., NAPCReN [19]) whenever possible. However, we also used a single instance of a certain ICD code across a number of databases to identify the presence of a comorbidity or history of a complication. Uncertainty of the comorbidity data in the EMR arose because information can be recorded in the problem list or encounter table and not always both. We acknowledge that comorbidities and complications may be over- or under captured based on the definitions used.

Could not be answered

Duration of diabetes

Data regarding the duration of diabetes could not be extracted as there was no reliable data field available to answer this question. Accuracy is questionable as the EMR defaults to record the date when entry of diagnosis was put into the EMR if no date is specified. Historical data from the previous EMR was not merged into the new EMR when it was adopted by all clinics in 2017. As well, this variable is primarily completed via patient-physician conversations and is often not documented in the extractable data field. Finally, external data sources do not feed into the EMR to populate this field.

New versus follow-up visits

No consistent indication/coding in EMR.

Proportion of individuals with hyperglycemic hyperosmolar state (HHS)

HHS diagnosis is not captured within the Alberta Health ICD coding scheme.