Patient and Health Care Factors Associated With Long-term Diabetes Complications Among Adults With and Without Mental Health and Substance Use Disorders

Key Points Question Among patients with newly diagnosed diabetes, are preexisting mental health (MH) or substance use (SU) disorders and primary care utilization before a new diabetes diagnosis associated with the long-term severity of diabetes complications? Findings In this cohort study of 122 992 patients in the US Department of Veterans Affairs health system, more than 90% of patients with MH or SU disorders had primary care visits before diabetes was newly diagnosed, compared with approximately 58% of patients without MH or SU disorders. Patients with MH or SU disorders experienced significantly lower severity of diabetes complications for 7 years thereafter, compared with patients without preexisting MH or SU disorders, even after controlling for sociodemographic characteristics and medical comorbidities. Meaning Among patients with MH or SU disorders, receiving more medical and mental health care from an integrated health care system before the onset of diabetes is associated with modest, albeit impermanent, health benefits after the onset of diabetes.


Index
Year 1  The level of missing outcome data in our analysis should be interpreted in the context of the level of each patient's data completeness as well. We used mixed-effects linear regression to run a longitudinal growth model with 8 total time points during which the outcome variable was calculated. That is, each patient's medical record data could lead to a maximum of 8 outcome data points observed over time in our analyses. Among the cohort as a whole, a technical majority of patients had at least 1 missing outcome data point at during a time period in our analyses (median outcome data points per patients = 7). However, for the main analysis, an average of 6.5 (SD = 2.1) outcome data points were analyzed per patient, indicating that most patients had many outcome data points over time, even if that patient had missing data during any given year or years.

Rates of Specific Diabetes Complications and Severity Over Time
This section of supplemental material first provides illustrative examples of the specific diagnoses for diabetes complications that were grouped, and weighted, to calculate a patient's score on the Diabetes Complication Severity Index (DCSI). Note that the method for calculating the DCSI was developed and described fully by Young et al. 1 Briefly, we analyzed patient's medical record to identify each diagnosis in the DCSI representing a medical abnormality consistent with diabetes complications (e.g., diabetic ophthalmologic disease). Each diagnosis we identified was assigned a weight of 1 or 2, per Young et al.'s specifications. 1 Within a category of complications such as Retinopathy or Nephropathy (i.e., the italicized rows in eTable3), the sum of weighted scores was calculated for a within-category total. The maximum within-category total allowed was 2 points, except for the neuropathy category which was only weighted at 1 point maximum (

Illustrative Examples of Diagnoses Included in Group Metabolic Complications
Metabolic Complications (weighted score = 2 for all) Ketoacidosis Hyperosmolar Other coma eTable 4 contains the percent of patients who received a diagnosis for for each category of complications in the DCSI, by comparative MH or SU disorder groups. In a given time period, each patient could have had both a mild and a severe complication within a given category. As such, column totals in eTable 4 may not total to 100%, nor will a row total to the "any" complication data presented in Table 2 of the manuscript. Table 2 of the manuscript displays rates of patients' highest-weighted within-category score on the DCSI across all 8 time periods. In eTable 4 of the Supplement, we present full data on event rates for the outcome variable, by comparative group, and by the 8 time periods analyzed in mixed-effects regressions.
Note also that we provide percentages in eTable 4 based on the number of patients who were alive in each time period (denominators are provided in the top rows of each respective section of the table).

Complete Mixed-Effects Regression Analysis Results
This section of the Supplement provides complete mixed-effects regression results for the main analysis and for both sensitivity analyses.