A survey of Alberta pharmacists' actions and opinions in regard to administering vaccines and medications by injection.

BACKGROUND: Pharmacists in Alberta have been authorized to administer vaccines and other medications by injection for more than 10 years; however, little is known about the provision of this service and their opinions regarding this service. Understanding pharmacists' experiences regarding injection services would inform development of strategies to improve provision of injection services. OBJECTIVES: To describe the actions related to administering an injection, including identification of commonly administered medications, and to identify perceived barriers and facilitators pharmacists face when providing injection services. METHODS: An online survey was developed and loaded into REDCap, and e-mail invitations were sent to 5714 pharmacists registered with the Alberta College of Pharmacy in October 2020. Responses were analyzed using descriptive statistics. Pharmacists who administered at least one injection in the previous year were considered active providers, and their opinions regarding injection services were compared with nonactive providers. RESULTS: A total of 397 pharmacists responded to our survey, mean age was 42 years, 66% were female, 82% were community pharmacists, and 90% were active providers. The most common injection, administered by 98% of active providers, was influenza vaccine, followed by vitamin B12 (95%), herpes zoster vaccine (88%), hepatitis vaccines (86%), and pneumococcal vaccines (82%). Nonactive providers were more likely than active providers to report that comfort with administering injections (P < 0.001) and managing adverse reactions (P = 0.013) were moderate or major barriers to providing injections. More than 60% of pharmacists indicated that access and automated reporting to the provincial immunization registry would be essential to increasing the frequency of providing injection services. CONCLUSION: We identified that Alberta pharmacists administer a wide variety of vaccines and other medications by injection. Respondents identified several barriers and facilitators to providing these services. Addressing these barriers may help improve provision of injection services by pharmacists.

Predicting 30-Day Readmissions in Patients With Heart Failure Using Administrative Data: A Machine Learning Approach.

BACKGROUND: We sought to develop machine learning (ML) models trained on administrative data which predict risk of readmission in patients with heart failure and to evaluate and compare the ML model with the currently used LaCE score using clinically informative metrics. METHODS AND RESULTS: This prognostic study was conducted in Alberta, Canada, on 9845 patients with confirmed heart failure admitted to hospital between 2012 and 2019. The outcome was unplanned all-cause hospital readmission within 30 days of discharge. We used 80% of the data for the ML model development and 20% for independent validation. We reported, using the validation set, c-statistics (area under the receiver operating characteristic curves)and performance metrics (likelihood ratio, positive predictive values) for the XGBoost model and a modified LaCE score within their respective predictive thresholds. Boosted tree-based classifiers had higher area under the receiver operating characteristic curves (0.65 for XGBoost) compared with others (0.58 for neural networks) and 0.57 for the modified LaCE. Within the predicted threshold range of the XGBoost classifier, the positive likelihood ratio was 1.00 at the low end of predicted risk and 6.12 at the high end, resulting in a positive predictive value (post-test probability) range of 21%-62%; the pretest probability of readmission was 20.9% using prevalence. The corresponding positive likelihood ratios and positive predictive values across LaCE score thresholds were 1.00-1.20 and 21%-24%, respectively. CONCLUSIONS: Despite predicting readmissions better than the LaCE, even the best ML model trained on administrative health data (XGBoost) did not provide substantially informative prediction performance as it only generated a moderate shift from pre to post-test probability. Health systems wishing to deploy such a tool should consider training ML models with additional data. Adding other techniques like natural language processing, along with ML, to use other clinical information (like chart notes) might improve prediction performance.

Depression and antecedent medication adherence in a cohort of new metformin users.

AIMS: The association between depression and poor medication adherence is based on cross-sectional studies and cohort studies that measure adherence rates after depression status is determined. However, depressive symptoms occur well before diagnosis. This study examined adherence patterns in the year before a depressive episode. METHODS: This retrospective cohort study followed new metformin users identified in Alberta Health's administrative data between 2008 and 2018. Depressive episodes starting ≥1 year after metformin initiation were identified using a validated case definition. Controls were randomly assigned a pseudo depression date. Adherence to oral antihyperglycemic medications was estimated using proportion of days covered (PDC) and group-based trajectory models to explore the association between depression and poor adherence (PDC<0.8). RESULTS: A depressive episode occurred in 17,418 (10.6%) of 165,056 new metformin users. Individuals with depression were more likely to have poor adherence compared to controls (adjusted odds ratio 1.21; 95% CI 1.17, 1.26). Five trajectories were identified: nearly perfect adherence (PDC >0.95 [34.8% of cohort]), discontinued (PDC=0 [18.3% of cohort], poor initial adherence (PDC 0.75) that declined either rapidly (9.2% of cohort) or gradually (30.1% of cohort), and poor initial adherence (PDC 0.26) that increased gradually (7.6% of cohort). Individuals with depression were more likely to be in one of the four trajectories of poor adherence compared to controls (adjusted odds ratio 1.24; 95% CI 1.19-1.29). CONCLUSIONS: Poor medication adherence occurs in the year before a depressive episode; therefore, poor medication use patterns could be used as an early warning sign for depression.

Limited Effectiveness of Diabetes Risk Assessment Tools in Seniors' Facility Residents.

BACKGROUND: Undiagnosed diabetes can create significant management issues for seniors. OBJECTIVES: To evaluate the effectiveness of two diabetes risk surveys-the Canadian Diabetes Risk Assessment Questionnaire (CANRISK) and the Finnish Diabetes Risk Score (FINDRISC)-to identify elevated blood glucose levels in seniors. METHODS: A cross-sectional study was conducted in senior living facilities in Edmonton, Alberta, Canada. Those with known diabetes, without capacity, considered frail, or unable to communicate in English were excluded. Participants completed the CANRISK and FINDRISC surveys and had their glycated hemoglobin A1c (HbA1c) measured. Correlations between seniors with elevated risk on the surveys and an HbA1c value of 6.5% or higher or 6.0% and higher were assessed. RESULTS: In this study, 290 residents participated; their mean age was 84.3 ± 7.3 years, 82 (28%) were men, and their mean HbA1c level was 5.7% ± 0.4%. Mean CANRISK score was 29.4 ± 8.0, and of the 254 (88%) considered to be moderate or high risk, 10 (4%) had an HbA1c level of 6.5% or higher and 49 (19%) had an HbA1c level of 6.0% or higher. Mean FINDRISC score was 10.8 ± 4.2, and of the 58 (20%) considered to be high or very high risk, 4 (7%) had an HbA1c level of 6.5% or higher and 15 (26%) had an HbA1c level of 6.0% or higher. The area under the receiver-operating characteristic curve was 0.57 (95% confidence interval 0.42-0.72) for the CANRISK survey identifying participants with an HbA1c level of 6.5% or higher and 0.59 (95% confidence interval 0.51-0.67) for identifying participants with an HbA1c level of 6.0% or higher. Similar characteristics were observed for the FINDRISC survey. CONCLUSIONS: In this group of seniors with no known diabetes history, mean HbA1c level approximated that in the general population and neither survey effectively identified those with elevated blood glucose levels. These findings should be confirmed in a larger study; nevertheless, routine use of these surveys as a diabetes screening strategy does not appear to be warranted at this time.

Community Pharmacy-Based Inducement Programs Associated With Better Medication Adherence: A Cohort Study.

BACKGROUND: Inducement programs can promote customer loyalty; however, the clinical effects of these programs are unknown. OBJECTIVE: To examine relationships among inducement program use, medication adherence, and health outcomes. METHODS: Alberta residents with ≥ 1 physician visit for diabetes or hypertension between April 2008 and March 2014 were eligible for this study and included if they were new statin users and alive at least 455 days after the first statin dispensation. Group assignment was based on whether all statin dispensations in the first year were obtained from pharmacies with or without inducement programs. Discontinuation was defined as no statin dispensations between 275 and 455 days after the first statin dispensation. Acute coronary syndrome (ACS) hospitalizations or deaths were identified between 456 days and 3 years after the first statin dispensation. Multivariable regression analyses were conducted to examine relationships among inducement program use, discontinuation, and ACS events. RESULTS: Among the 159 998 new statin users, mean age was 60.2 (±13.7) years and 67 534 (42%) were women. Statin discontinuation occurred in 22 455 (28.9%) of 77 803 inducement group participants and 25 816 (31.4%) of 82 195 noninducement group participants (adjusted odds ratio = 0.88; 95% CI = 0.86-0.90). Risk of an ACS event was similar between groups (adjusted hazard ratio = 1.00; 95% CI 0.92-1.08); however, discontinuing statin therapy was associated with a higher risk of an ACS event (adjusted hazard ratio = 1.27; 95% CI = 1.16-1.39). CONCLUSIONS: Inducement programs are associated with better adherence and not directly associated with risk of health outcomes.

Medication Adherence Affects Risk of New Diabetes Complications: A Cohort Study.

BACKGROUND: Previous outcomes-based studies of adherence to diabetes medications have focused on glycemic control and are limited by questions of temporality and uncontrolled confounding. OBJECTIVE: This retrospective cohort study of new oral antidiabetic medication users examined the effect of adherence on risk of incident macrovascular and microvascular complications. METHODS: A nationwide integrated insurance claims and laboratory database was used to identify new oral antidiabetic medication users between January 2004 and December 2009. People with preexisting complications were excluded and the remaining cohort was followed until development of a new diabetes complication or December 2010. Medication adherence was calculated at 3-month intervals and entered as a time-dependent variable in a Cox proportional hazards model. Covariables entered in the model included patient demographics, clinical laboratory data, a medical frailty indicator and a mortality risk score from the Johns Hopkins adjusted clinical groups system, and medication use at baseline. RESULTS: Among the 54 505 included patients, the median age was 60 years, 28 125 (52%) were men, 1447 (3%) were considered frail, the mean mortality risk score was 33.7 (±11.1), and 9793 (18%) developed a new diabetes complication. Good adherence (medication possession ratio ≥0.8) was associated with a lower risk of a new microvascular or macrovascular diabetes complication (adjusted hazard ratio = 0.96; 95% CI = 0.92-1.00; P = 0.05). CONCLUSIONS: This study design addresses limitations of previous studies and found a small but significantly lower risk of new diabetes complications associated with good adherence to oral antidiabetic medications.

Dose-response relationship between sulfonylureas and major adverse cardiovascular events in elderly patients with type 2 diabetes.

PURPOSE: The objective of this study was to determine if there is a dose-response relationship between sulfonylureas and major adverse cardiovascular events (MACE). METHODS: We conducted a retrospective cohort study among elderly patients with no history of acute coronary syndrome or stroke who initiated gliclazide or glyburide therapy between 1998 and 2010. Gliclazide and glyburide users were evaluated separately, and a high-dimensional propensity score (HDPS) was used to match patients initiating therapy with a low or high dose. A time-dependent variable was used to further characterize exposure, which can change during follow-up. Cox proportional hazard regression models were used to compare the risk of MACE between low (reference) and high doses. RESULTS: We identified 14,213 new users of gliclazide or glyburide (mean age, 74.7 (standard deviation 6.4) years; males, 52.8%; and mean follow-up duration, 2.7 (standard deviation 2.9) years). Among gliclazide users, there was a higher risk of MACE with high compared with low dose (crude rates: 32.8 and 28.2 per 1000 person-years, respectively), but this did not reach statistical significance (HDPS-matched hazard ratio) 1.15; 95% confidence interval (0.96-1.38). For glyburide users, however, MACE occurred more frequently in the high compared with low dose (crude rates: 38.9 and 31.5 per 1000 person-years, respectively; HDPS-matched hazard ratio 1.24; 95% confidence interval 1.02-1.50). CONCLUSIONS: Among new users of sulfonylureas, there appears to be a dose-response relationship between glyburide and MACE, but not for gliclazide. Copyright © 2016 John Wiley & Sons, Ltd.

Pharmacists on primary care teams: Effect on antihypertensive medication management in patients with type 2 diabetes.

OBJECTIVE: To identify which activities produced a significant improvement in blood pressure control in patients with type 2 diabetes when pharmacists were added to primary care teams. METHODS: This prespecified, secondary analysis evaluated medication management data from a randomized controlled trial. The primary outcome was a change in treatment, defined as addition, dosage increase, or switching of an antihypertensive medication during the 1-year study period. The secondary outcome was a change in antihypertensive medication adherence using the medication possession ratio (MPR). RESULTS: The 200 evaluable trial patients had a mean age of 59 (SD, 11) years, 44% were men, and mean blood pressure was 130 (SD, 16)/74 (SD, 10) mm Hg at baseline. Treatment changes occurred in 45 (42%) of 107 patients in the intervention group and 24 (26%) of 93 patients in the control group (RR, 1.63; 95% CI, 1.08-2.46). Addition of a new medication was the most common type of change, occurring in 34 (32%) patients in the intervention group and 17 (18%) patients in the control group (P = 0.029). Adherence to antihypertensive medication was high at baseline (MPR, 93%). Although medication adherence improved in the intervention group (MPR, 97%) and declined in the control group (MPR, 91%), the difference between groups was not significant (P = 0.21). CONCLUSION: The observed improvement in blood pressure control when pharmacists were added to primary care teams was likely achieved through antihypertensive treatment changes and not through improvements in antihypertensive medication adherence.

Changes in thiazolidinedione use and outcomes following removal of a prior-authorization policy: controlled time-series analysis.

OBJECTIVE: The aim of this study was to assess the impact of removing prior-authorization restrictions on the use of thiazolidinediones (TZDs) and outcomes. METHODS: In a controlled interrupted time-series analysis, whereby new users of antidiabetic agents over 65 years of age in adjacent Canadian provinces with different TZD-related policies [Alberta (n = 16,653) and Saskatchewan (n = 6682)] were followed from January 2001 to December 2006. Prior authorization for TZDs was removed in Alberta (intervention province) in December 2003 (rosiglitazone) and February 2004 (pioglitazone); no policy changes occurred in Saskatchewan (control province). Adjusted differences in percent change between intervention and control provinces were used to estimate policy-attributable effects (PAE) on TZD use within 30 days and 1 year and patient outcomes (composite of all-cause mortality or hospitalization for acute coronary events or heart failure) within 1 year. RESULTS: Mean age was 75 years, 51% were female, and 206,055 antidiabetic prescriptions were dispensed during follow-up. TZD use within 30 days among new users of antidiabetic agents increased almost >10% in Alberta compared with Saskatchewan controls immediately following the policy change: PAE = 9.4%, 95% confidence interval, 7.3%-11.6%. Other less expensive antidiabetic drug use decreased to exactly the same extent, suggesting TZD substitution. Compared with the controls, in Alberta there were no changes in the primary (clinical) composite outcome at 1 year (PAE = 0.31%, 95% confidence interval, -2.8% to +3.4%). CONCLUSIONS: The removal of a prior-authorization policy for TZDs was associated with an immediate increase in TZD use but did not impact patient outcomes. In this case, the policy removal shifted drug use to a more expensive drug with less certain clinical benefit.

Are Guideline-concordant Processes of Care Consistent Across the Rural-Urban Continuum? A Retrospective Cohort Study of Adults Newly Treated for Type 2 Diabetes.

OBJECTIVES: Our aim in this study was to identify the association between place of residence (metropolitan, urban, rural) and guideline-concordant processes of care in the first year of type 2 diabetes management. METHODS: We conducted a retrospective cohort study of new metformin users between April 2015 and March 2020 in Alberta, Canada. Outcomes were identified as guideline-concordant processes of care through the review of clinical practice guidelines and published literature. Using multivariable logistic regression, the following outcomes were examined by place of residence: dispensation of a statin, angiotensin-converting enzyme inhibitor (ACEi) or angiotensin II receptor blocker (ARB), eye examination, glycated hemoglobin (A1C), cholesterol, and kidney function testing. RESULTS: Of 60,222 new metformin users, 67% resided in a metropolitan area, 10% in an urban area, and 23% in a rural area. After confounder adjustment, rural residents were less likely to have a statin dispensed (adjusted odds ratio [aOR] 0.83, 95% confidence interval [CI] 0.79 to 0.87) or undergo cholesterol testing (aOR 0.86, 95% CI 0.83 to 0.90) when compared with metropolitan residents. In contrast, rural residents were more likely to receive A1C and kidney function testing (aOR 1.14, 95% CI 1.08 to 1.21 and aOR 1.17, 95% CI 1.11 to 1.24, respectively). ACEi/ARB use and eye examinations were similar across place of residence. CONCLUSIONS: Processes of care varied by place of residence. Limited cholesterol management in rural areas is concerning because this may lead to increased cardiovascular outcomes.

Rurality is associated with lower likelihood of dipeptidyl peptidase 4 inhibitor use for treatment intensification.

Background: Antihyperglycemic drug utilization studies are conducted frequently and describe the uptake of new drug therapies across may jurisdictions. An increasingly important, yet often absent, aspect of these studies is the impact of rurality on drug utilization. Objectives: The objective of this study was to explore the association between place of residence (rural, urban, metropolitan) and the use of dipeptidyl peptidase 4 inhibitors (DPP-4i) for first treatment intensification of type 2 diabetes. Methods: A retrospective cohort study was conducted from April 1, 2008 to March 31, 2019 of new metformin users. A multivariable logistic regression analysis was performed to determine the association between place of residence (using postal codes) and likelihood of DPP-4i dispensing. Results: After adjusting for confounders, analysis revealed that rural-dwellers are less likely to have a DPP-4i dispensed, compared with metropolitan-dwellers (aOR:0.64; 95%CI:0.61-0.67) and over-time, the uptake in rural areas was slower. Conclusions: This study demonstrates that rurality can have an impact on drug therapy decisions at first treatment intensification, with respect to the utilization of new therapies.

Adding pharmacists to primary care teams increases guideline-concordant antiplatelet use in patients with type 2 diabetes: results from a randomized trial.

BACKGROUND: Antiplatelet therapy is recommended as part of a strategy to reduce the risk of cardiovascular events in patients with type 2 diabetes. However, compliance with these guideline-recommended therapies appears to be less than ideal. OBJECTIVE: To assess the effect of adding pharmacists to primary care teams on initiation of guideline-concordant antiplatelet therapy in type 2 diabetic patients. METHODS: Prespecified secondary analysis of randomized trial data. In the main study, the pharmacist intervention included a complete medication history, limited physical examination, provision of guideline-concordant recommendations to the physician to optimize drug therapy, and 1-year follow-up. Controls received usual care without pharmacist interactions. Patients with an indication for antiplatelet therapy, but not using an antiplatelet drug at randomization were included in this substudy. The primary outcome was the proportion of patients using an antiplatelet drug at 1 year. RESULTS: At randomization, 257 of 260 study patients had guideline-concordant indications for antiplatelet therapy, but less than half (121; 47%) were using an antiplatelet drug. Overall, 136 patients met inclusion criteria for the substudy (71 intervention and 65 controls): 60% were women, with mean (SD) age 58.0 (11.9) years, diabetes duration 5.3 (6.0) years, and hemoglobin A(1c) 7.6% (1.5). Sixteen (12%) had established cardiovascular disease at enrollment. At 1 year, 43 (61%) intervention patients and 15 (23%) controls were using an antiplatelet drug (38% absolute difference; number needed to treat, 3; relative increase, 2.6; 95% CI 1.5-4.7; p < 0.001). Of these 58 patients, 52 (90%) were using aspirin 81 mg daily. CONCLUSIONS: Adding pharmacists to primary care teams significantly and substantially increased the proportion of type 2 diabetic patients using guideline-concordant antiplatelet therapy.

Evaluating the introduction of a computerized prior-authorization system on the completeness of drug exposure data.

PURPOSE: Administrative databases that only capture records for benefit-approved prescriptions may underestimate exposure because they do not capture non-benefit prescriptions. Using a natural experiment, we illustrate the impact of automating a prior-authorization policy on the completeness of drug exposure. METHODS: Using Saskatchewan (Canada) databases, weekly counts of benefit-approved and total prescription records in 2006 for new users of antidiabetic agents were examined across four categories: thiazolidinediones (TZDs), metformin, glyburide, and insulin. On July 1, 2006, Saskatchewan's public drug plan implemented an automated, online-adjudicated, prior-authorization process for TZDs; previously, prior approval was paper based. No such policy changes occurred for other drugs. We estimated the effect of this policy change on drug exposure using interrupted time-series analyses. RESULTS: We examined 223 552 prescription records: 19% were for TZDs, 48% for metformin, 20% for glyburide, and 13% for insulin. Prior to automation, there were, on average, 571 benefit-approved TZD records per week; however, the number of benefit-approved TZD records increased immediately after the automated process was introduced by 240 prescriptions per week (95% CI 200-280, p < 0.001). The average proportion of TZD benefit-approved records was 73% before and increased to 93% immediately following policy change (20% absolute change, 95% CI 18.7-20.4%). No changes were observed for metformin, glyburide, or insulin (p > 0.1 for all). CONCLUSIONS: Automating prior authorization for TZDs immediately increased the proportion of captured TZD records, suggesting in our study that one-fifth of TZD exposure was previously misclassified. If replicable, this indicates that even subtle changes in reimbursement policy may affect the validity of drug exposure data.

Rural Residence Is Associated With a Delayed Trend Away From Sulfonylurea Use for Treatment Intensification of Type 2 Diabetes.

OBJECTIVE: To examine the intersection between location of residence along the rural-urban continuum (metropolitan, urban, and rural) and sulfonylurea dispensation records for the management of type 2 diabetes. RESEARCH DESIGN AND METHODS: This retrospective cohort study used administrative health records of adult new metformin users between April 2008 and March 2019 in Alberta, Canada. Multivariable logistic regression was performed to examine the association between sulfonylurea-based treatment intensification and location of residence. RESULTS: Treatment was intensified in 66,084 (38%) of 171,759 new metformin users after a mean of 1.5 years. At treatment intensification, mean age was 55 years, 62% of users were male, and 27% were rural residents. The most common antihyperglycemic drug, given to 30,297 people (46%) for treatment intensification, was a sulfonylurea. At the beginning of our observation period, the proportion of people dispensed a sulfonylurea at first treatment intensification was highest in rural (57%), compared with urban (54%) and metropolitan (52%) areas (P = 0.009). Although proportions decreased over time across the province, rural residents continued to constitute the highest proportion of sulfonylurea users (45%), compared with urban (35%) and metropolitan (37%) residents (P < 0.001), and the trend away from sulfonylurea use was delayed by ∼4 years for rural residents. Adjusting for potential sources of confounding, rural residence was associated with a significantly higher likelihood of using a sulfonylurea compared with metropolitan residence (adjusted odds ratio 1.34; 95% CI 1.29-1.39). CONCLUSIONS: Variation in sulfonylurea dispensation across the rural-urban continuum provides a basis for continued research in the differences in process of care by location.

Predicting 30-day risk from benzodiazepine/Z-drug dispensations in older adults using administrative data: A prognostic machine learning approach.

OBJECTIVE: To develop a machine-learning (ML) model using administrative data to estimate risk of adverse outcomes within 30-days of a benzodiazepine (BZRA) dispensation in older adults for use by health departments/regulators. DESIGN, SETTING AND PARTICIPANTS: This study was conducted in Alberta, Canada during 2018-2019 in Albertans 65 years of age and older. Those with any history of malignancy or palliative care were excluded. EXPOSURE: Each BZRA dispensation from a community pharmacy served as the unit of analysis. MAIN OUTCOMES AND MEASURES: ML algorithms were developed on 2018 administrative data to predict risk of any-cause hospitalization, emergency department visit or death within 30-days of a BZRA dispensation. Validation on 2019 administrative data was done using XGBoost to evaluate discrimination, calibration and other relevant metrics on ranked predictions. Daily and quarterly predictions were simulated on 2019 data. RESULTS: 65,063 study participants were included which represented 633,333 BZRA dispensation during 2018-2019. The validation set had 314,615 dispensations linked to 55,928 all-cause outcomes representing a pre-test probability of 17.8%. C-statistic for the XGBoost model was 0.75. Measuring risk at the end of 2019, the top 0.1 percentile of predicted risk had a LR + of 40.31 translating to a post-test probability of 90%. Daily and quarterly classification simulations resulted in uninformative predictions with positive likelihood ratios less than 10 in all risk prediction categories. Previous history of admissions was ranked highest in variable importance. CONCLUSION: Developing ML models using only administrative health data may not provide health regulators with sufficient informative predictions to use as decision aids for potential interventions, especially if considering daily or quarterly classifications of BZRA risks in older adults. ML models may be informative for this context if yearly classifications are preferred. Health regulators should have access to other types of data to improve ML prediction.

Predicting falls-related admissions in older adults in Alberta, Canada: a machine-learning falls prevention tool developed using population administrative health data.

OBJECTIVE: To construct a machine-learning (ML) model for health systems with organised falls prevention programmes to identify older adults at risk for fall-related admissions. DESIGN: This prognostic study used population-level administrative health data to develop an ML prediction model. SETTING: This study took place in Alberta, Canada during 2018-2019. PARTICIPANTS: Albertans aged 65 and older with at least one prior admission. Those with palliative conditions or emigrated out of Alberta were excluded. EXPOSURE: Unit of analysis was the individual person. MAIN OUTCOMES/MEASURES: We identified fall-related admissions. A CatBoost model was developed on 2018 data to predict risk of fall-related emergency department visits or hospitalisations. Temporal validation was done using 2019 data to evaluate model performance. We reported discrimination, calibration and other relevant metrics measured at the end of 2019 on both ranked predictions and predicted probability thresholds. A cost-savings simulation was performed using 2019 data. RESULTS: Final number of study participants was 224 445. The validation set had 203 584 participants with 19 389 fall-related events (9.5% pretest probability) and an ML model c-statistic of 0.70. The highest ranked predictions had post-test probabilities ranging from 40% to 50%. Net benefit analysis presented mixed results with some net benefit using the ML model in the 6%-30% range. The top 50 percentile of predicted risks represented nearly $C60 million in health system costs related to falls. Intervening on the top 25 or 50 percentiles of predicted risk could realise substantial (up to $C16 million) savings. CONCLUSION: ML prediction models based on population-level administrative data can assist health systems with fall prevention programmes identify older adults at risk of fall-related admissions and reduce costs. ML predictions based on ranked predictions or probability thresholds could guide subsequent interventions to mitigate fall risks. Increased access to diverse forms of data could improve ML performance and further reduce costs.

Foregone Opportunities? Time to Treatment Intensification Among Adults With Type 2 Diabetes and Elevated Glycated Hemoglobin on Metformin Monotherapy, 2009-2018.

OBJECTIVES: In this study, we aimed to characterize time to treatment intensification (TTI) in people on metformin with uncontrolled hyperglycemia, and estimated the frequency of physician visits until intensification. METHODS: This work was a cohort study of Albertan adults with glycated hemoglobin (A1C) of >7.5% after at least 3 months of metformin monotherapy, using administrative databases from 2009 to 2018, with each subject followed for up to 4 years. Therapeutic intensification was defined as dispensation of an additional class of antihyperglycemic medication. Median TTI and the median number of physician visits were estimated from Kaplan-Meier functions within age/A1C strata. A Cox proportional hazards model was fitted to examine predictors of therapeutic intensification. RESULTS: We included 38,846 people (average age, 57 years; 37% female; mean A1C, 8.8%). Overall, therapeutic intensification was observed in 23,077 (59%; 40% at 1 year). Median TTI was 1.4 years, varying from 0.7 years (A1C >8.5%, age <65 years) to 3.3 years (age ≥75 years, any A1C). The median number of physician visits until intensification was 9, varying between 5 (A1C >8.5%, age <65 years) and ≥30 (age ≥75 years); 93% of people awaiting intensification had at least 2 visits by 1 year. Higher A1C and younger age were the strongest predictors of intensification. Results were similar in people with ischemic heart disease. CONCLUSIONS: Despite ample contacts with community physicians, TTI exceeds the 6-month target recommended by guidelines, particularly in older adults. Further study is needed to better understand these foregone opportunities as guidelines call for wider promulgation of agents with cardiorenal benefits.

Association Between Antidepressant Use and Adherence to Anti-hyperglycemic Medications in Adults With Type 2 Diabetes and Depression: A Retrospective Cohort Study.

OBJECTIVES: Depression is a known risk factor for poor medication adherence, but it is unclear whether depression treatment affects adherence rates. In this study, we examined the association between pharmacologic treatment of a new depressive episode and subsequent adherence to oral anti-hyperglycemic medications. METHODS: In this retrospective cohort study we used administrative health data to follow adult new metformin users in Alberta, Canada, between 2008 and 2018. Depressive episodes starting ≥1 year after metformin initiation were identified and individuals starting antidepressant treatment within the first 90 days were compared with those who did not. The proportion of days covered (PDC) with oral anti-hyperglycemic medications in the subsequent year (days 91 to 455) was used to estimate adherence. The association between antidepressant treatment and poor adherence (PDC<0.8) was examined using multivariate logistic regression models. RESULTS: A new depressive episode occurred in 6,201 people, with a mean age of 56.0 (standard deviation [SD], 15.4) years. Of this cohort, 3,303 (53.2%) were women. Mean PDC was 0.55 (SD, 0.41); 924 (57.0%) of 1,621 people who started antidepressant treatment and 2,709 (59.2%) of 4,580 controls had poor adherence (p=0.13). After adjusting for baseline comorbidities and other characteristics, antidepressant treatment was associated with a lower likelihood of poor adherence (adjusted odds ratio, 0.85; 95% confidence interval, 0.75 to 0.96; p=0.007). CONCLUSIONS: Although overall adherence to anti-hyperglycemic medications was low after onset of a depressive episode, antidepressant treatment was associated with a lower likelihood of poor adherence.

Development and Validation of a Machine Learning Model to Estimate Risk of Adverse Outcomes Within 30 Days of Opioid Dispensation.

Importance: Machine learning approaches can assist opioid stewardship by identifying high-risk opioid prescribing for potential interventions. Objective: To develop a machine learning model for deployment that can estimate the risk of adverse outcomes within 30 days of an opioid dispensation as a potential component of prescription drug monitoring programs using access to real-world data. Design, Setting, and Participants: This prognostic study used population-level administrative health data to construct a machine learning model. This study took place in Alberta, Canada (from January 1, 2018, to December 31, 2019), and included all patients 18 years and older who received at least 1 opioid dispensation from a community pharmacy within the province. Exposures: Each opioid dispensation served as the unit of analysis. Main Outcomes and Measures: Opioid-related adverse outcomes were identified from administrative data sets. An XGBoost model was developed on 2018 data to estimate the risk of hospitalization, an emergency department visit, or mortality within 30 days of an opioid dispensation; validation on 2019 data was done to evaluate model performance. Model discrimination, calibration, and other relevant metrics are reported using daily and weekly predictions on both ranked predictions and predicted probability thresholds using all data from 2019. Results: A total of 853 324 participants represented 6 181 025 opioid dispensations, with 145 016 outcome events reported (2.3%); 46.4% of the participants were men and 53.6% were women, with a mean (SD) age of 49.1 (15.6) years for men and 51.0 (18.0) years for women. Of the outcome events, 77 326 (2.6% pretest probability) occurred within 30 days of a dispensation in the validation set (XGBoost C statistic, 0.82 [95% CI, 0.81-0.82]). The top 0.1 percentile of estimated risk had a positive likelihood ratio (LR) of 28.7, which translated to a posttest probability of 43.1%. In our simulations, the weekly measured predictions had higher positive LRs in both the highest-risk dispensations and percentiles of estimated risk compared with predictions measured daily. Net benefit analysis showed that using machine learning prediction may not add additional benefit over the entire range of probability thresholds. Conclusions and Relevance: These findings suggest that prescription drug monitoring programs can use machine learning classifiers to identify patients at risk of opioid-related adverse outcomes and intervene on high-risk ranked predictions. Better access to available administrative and clinical data could improve the prediction performance of machine learning classifiers and thus expand opioid stewardship efforts.

Effect of aspirin dose on mortality and cardiovascular events in people with diabetes: a meta-analysis.

OBJECTIVES: Pharmacologic evidence suggests adequate antiplatelet activity in diabetic patients requires >100 mg aspirin daily, yet recent trials have used ≤100 mg daily. This meta-analysis explored the relationship between aspirin dose and prevention of cardiovascular events. DATA SOURCES: Six electronic databases were searched using database-appropriate terms for aspirin, diabetes, and comparative study from inception until February 2010. REVIEW METHODS: Randomized controlled trials and cohort studies comparing aspirin to no antiplatelet therapy were included if they reported cardiovascular events as pre-specified outcomes, aspirin dose, and number of diabetic patients. Studies were stratified by daily aspirin dose (≤100 mg; 101-325 mg; >325 mg) and pooled risk ratios (RR) were calculated using random effects models. All-cause mortality was the primary outcome of interest. Cardiovascular-related mortality, myocardial infarction, and stroke were secondary outcomes. RESULTS: Data for diabetic patients were available from 21 studies (n = 17,522). Overall, 1,172 (15.4%) of 7,592 aspirin users and 1,520 (18.4%) of 8,269 controls died (p = 0.31). The pooled RRs were 0.89 (95% CI: 0.72-1.10; p = 0.27) from 13 studies using ≤100 mg (I(2) = 64%); 0.89 (95% CI: 0.61-1.30; p = 0.55) from four studies using 101-325 mg (I(2) = 83%); and 0.96 (95% CI: 0.85-1.08; p = 0.50) from eight studies using >325 mg (I(2) = 0%). Aspirin use was associated with a significantly lower risk of mortality (RR: 0.82; 95% CI: 0.69-0.98; p = 0.03) in 13 secondary prevention studies (I(2) = 27%), whereas aspirin use in seven primary prevention studies (I(2) = 0%) was not (RR: 1.01; 95% CI 0.85-1.19; p = 0.94). A substantial amount of heterogeneity was observed amongst studies in all outcomes. Although inclusion of cohort studies was a major source of heterogeneity, stratification by study design did not reveal a significant dose-response relationship. CONCLUSIONS/INTERPRETATION: This summary of available data does not support an aspirin dose-response effect for prevention of cardiovascular events in diabetic patients. However, the systematic review identified an important gap in randomized controlled trial evidence for using 101-325 mg aspirin daily in diabetes.