A machine learning model to predict therapeutic inertia in type 2 diabetes using electronic health record data.

OBJECTIVE: To estimate the therapeutic inertia prevalence for patients with type 2 diabetes, develop and validate a machine learning model predicting therapeutic inertia, and determine the added predictive value of area-level social determinants of health (SDOH). METHODS: This prognostic study with a retrospective cohort design used OneFlorida data (linked electronic health records (EHRs) from 1240 practices/clinics in Florida). The study cohort included adults (aged ≥ 18) with type 2 diabetes, HbA1C ≥ 7% (53 mmol/mol), ≥one ambulatory visit, and ≥one antihyperglycemic medication prescribed (excluded patients prescribed insulin before HbA1C). The outcome was therapeutic inertia, defined as absence of treatment intensification within six months after HbA1C ≥ 7% (53 mmol/mol). The predictors were patient, provider, and healthcare system factors. Machine learning methods included gradient boosting machines (GBM), random forests (RF), elastic net (EN), and least absolute shrinkage and selection operator (LASSO). The DeLong test compared the discriminative ability (represented by C-statistics) between models. RESULTS: The cohort included 31,087 patients with type 2 diabetes (mean age = 58.89 (SD = 13.27) years, 50.50% male, 58.89% White). The therapeutic inertia prevalence was 39.80% among the 68,445 records. GBM outperformed (C-statistic from testing sample = 0.84, 95% CI = 0.83-0.84) RF (C-statistic = 0.80, 95% CI = 0.79-0.80), EN (C-statistic = 0.80, 95% CI = 0.80-0.81), and LASSO (C-statistic = 0.80, 95% CI = 0.80-0.81), p < 0.05. Area-level SDOH significantly increased the discriminative ability versus models without SDOH (C-statistic for GBM = 0.84, 95% CI = 0.84-0.85 vs. 0.84, 95% CI = 0.83-0.84), p < 0.05. CONCLUSIONS: Using EHRs of patients with type 2 diabetes from a large state, machine learning predicted therapeutic inertia (prevalence = 40%). The model's ability to predict patients at high risk of therapeutic inertia is clinically applicable to diabetes care.

Lack of evidence that beta blocker use reduces knee pain, areas of joint pain, or analgesic use among individuals with symptomatic knee osteoarthritis.

OBJECTIVE: The potential for beta blocker use to reduce joint pain and analgesic use in osteoarthritis (OA) patients has not been well established. The objective of this study was to estimate the association between beta blocker use and knee pain, areas of joint pain, and analgesic use among participants with symptomatic knee OA. DESIGN: We selected participants with symptomatic knee OA from the Osteoarthritis Initiative. Outcome measures included knee pain (e.g., WOMAC pain subscale), areas of joint pain (e.g., widespread joint pain), and analgesic use (e.g., use of strong pain prescriptions including opioids). We decomposed time-varying beta blocker use into within-person and between-person variation, and included these components in linear mixed effects models for repeated outcome measures of knee pain, joint pain, and analgesic use over 8 years. RESULTS: Among 1,168 participants, 15% reported beta blocker use at baseline. Beta blocker users (5.2, 95% CI [4.7, 5.8]) had similar estimated mean WOMAC pain scores as other anti-hypertensive users (4.9, 95% CI [4.6, 5.2]), with an estimated within-person difference of 0.1 (95% CI [-0.3, 0.4]). Proportion of participants reporting widespread joint pain was similar between beta blocker users and other anti-hypertensive users (40.1% vs 40.3%; within-person effect, odds ratio [OR] = 0.87, 95% CI [0.63, 1.22]). Reported use of strong prescription pain medication was also similar between beta blocker users and other anti-hypertensive users (7.7% vs 8.2%; within-person effect, OR = 1.39, 95% CI [0.75, 2.55]). CONCLUSIONS: We found no evidence that beta blockers confer a clinically meaningful reduction in knee pain severity, areas of joint pain, or analgesic use among participants with symptomatic knee OA.

Analgesic use and risk of recurrent falls in participants with or at risk of knee osteoarthritis: data from the Osteoarthritis Initiative.

OBJECTIVE: Few studies have compared the risk of recurrent falls across different types of analgesic use, and with limited adjustment for potential confounders (e.g., pain/depression severity). We assessed analgesic use and the subsequent risk of recurrent falls, among participants with or at risk of knee osteoarthritis (OA). METHODS: A longitudinal analysis included 4231 participants aged 45-79 years at baseline with 4-year follow-up from the Osteoarthritis Initiative (OAI) cohort study. We grouped participants into six mutually exclusive subgroups based on annually assessed analgesic use in the following hierarchical order of analgesic/central nervous system (CNS) potency: use of (1) opioids, (2) antidepressants, (3) other prescription pain medications, (4) over-the-counter (OTC) pain medications, (5) nutraceuticals, and (6) no analgesics. We used multivariable modified Poisson regression models with a robust error variance to estimate the effect of analgesic use on the risk of recurrent falls (≥2) in the following year, adjusted for demographics and health status/behavior factors. RESULTS: Opioid use increased from 2.7% at baseline to 3.6% at the 36-month visit (>80% using other analgesics/nutraceuticals), while other prescription pain medication use decreased from 16.7% to 11.9% over this time period. Approximately 15% of participants reported recurrent falls. Compared to those not using analgesics, participants who used opioids and/or antidepressants had a 22-25% increased risk of recurrent falls (opioids: RRadjusted = 1.22, 95% CI = 1.04-1.45; antidepressants: RRadjusted = 1.25, 95% CI = 1.10-1.41). CONCLUSION: Participants with or at risk of knee OA who used opioids and antidepressants with/without other analgesics/nutraceuticals may have an increased risk of recurrent falls after adjusting for potential confounders. Use of opioids and antidepressants warrants caution.