Novel application of approaches to predicting medication adherence using medical claims data

Predictive Modeling of Factors Influencing Adherence to SGLT-2 Inhibitors in Ambulatory Care: Insights from Prescription Claims Data Analysis

Sodium-glucose cotransporter 2 inhibitors (SGLT2i) are novel oral anti-hyperglycemic drugs that demonstrate cardiovascular and metabolic benefits for patients with type 2 diabetes (T2D), heart failure (HF), and chronic kidney disease (CKD). There is limited knowledge of real-world data to predict adherence to SGLT-2i in an ambulatory setting. The study aims to predict SGLT-2i adherence in patients with T2D and/or HF and/or CKD by building a prediction model using electronic prescription claims data presented within EPIC datasets. This is a retrospective study of 174 adult patients prescribed SGLT-2i at UC San Diego Health ambulatory pharmacies between 1 January 2020 to 30 April 2021. Adherence was measured by the proportion of days covered (PDC). R packages were used to identify regression and non-linear regression predictive models to predict adherence. Age, gender, race/ethnicity, hemoglobin A1c, and insurance plan were included in the model. Diabetes control based on hemoglobin A1c (HbA1c) and the glomerular filtration rate (GFR) was also evaluated using Welch t-test with a p-value of 0.05. The best predictive model for measuring adherence was the simple decision tree. It had the highest area under the curve (AUC) of 74% and accuracy of 82%. The model accounted for 21 variables with the main node predictors, including glycated hemoglobin, age, gender, and insurance plan payment amount. The adherence rate was inversely proportional to HbA1c and directly proportional to the plan payment amount. As for secondary outcomes, HbA1c values from baseline till 90 days post-treatment duration were consistently higher in the non-compliant group: 7.4% vs. 9.6%, p < 0.001 for the PDC ≥ 0.80 and PDC < 0.80, respectively. Baseline eGFR was 55.18 mL/min/1.73m2 vs. 54.23 mL/min/m2 at 90 days. The mean eGFR at the end of the study (minimum of 90 days of treatment) was statistically different between the groups: 53.1 vs. 59.6 mL/min/1.73 m2, p < 0.001 for the PDC ≥ 0.80 and PDC < 0.80, respectively. Adherence predictive models will help clinicians to tailor regimens based on non-adherence risk scores.

Machine learning-based prediction of medication refill adherence among first-time insulin users with type 2 diabetes

AIMS

The prevalence of Type 2 Diabetes Mellitus (T2DM) is projected to be 7 % in 2030. Despite its need for long-term diabetes care, the adherence rate of injectable medications such as insulin is around 60 %, lower than the acceptable threshold of 80 %. This study aims to create classification models to predict insulin adherence among adult T2DM naïve insulin users.

METHODS

Clinical data were extracted from Taipei Medical University Clinical Research Database (TMUCRD) from January 1st, 2004 to December 30th, 2020. A patient was regarded as adherent if his/her medication possession ratio (MPR) was at least 80 %. Seven domains of predictors were created, including demographics, baseline medications, baseline comorbidities, baseline laboratory data, healthcare resource utilization, index insulins, and the concomitant non-insulin T2DM medications. We built two Xgboost models for internal and external testing respectively.

RESULTS

Using a cohort of 4134 patients from Taiwan, our model achieved the Area Under the curve of the Receiver Operating Characteristic (AUROC) of the internal test was 0.782 and the AUROC of the external test was 0.771. the SHAP (SHapley Additive exPlanations) value showed that the number of prescribed medications, the number of outpatient visits, and laboratory data were predictive of future insulin adherence.

CONCLUSIONS

This is the first study to predict adherence among adult naïve insulin users. The developed model is a potential clinical decision support tool to identify possible non-adherent patients for healthcare providers to design individualized education plans.

A dynamic analysis of medication adherence

BACKGROUND: Medication adherence is an important factor in maintaining and improving health, although adherence levels are often suboptimal. Previous studies have highlighted the importance of prior adherence behavior in understanding future adherence behaviors. OBJECTIVE: To improve understanding of adherence behavior and analyze the role of previous adherence in estimating the likelihood of future adherence for maintenance medications. METHODS: The adherence behaviors of 53,709 continuously enrolled individuals in employer-sponsored health plans were analyzed using a state-dependence framework (ie, adherence patterns in the past influence adherence in the future). This allowed for the estimation of the extent of carryover in adherence from one quarter to another while adjusting for observed and unobserved heterogeneity and enrollee characteristics. The role of the initial observation of adherence on the likelihood of future adherence was also analyzed. This study focuses on enrollee cohorts who filled prescriptions in 3 maintenance medication classes: lipid-lowering medications, antihypertensive medications, and oral antidiabetes medications. RESULTS: If an enrollee was adherent in the previous quarter, more than 80% of the time they remained adherent in the current quarter. Similarly, if they were nonadherent in the previous quarter, more than 75% of the time they remained nonadherent. Marginal effect estimates for prior adherence (previous quarter and initial quarter) showed increases in predicted adherence when adherent in the previous quarter (8.7 percentage points [pp] [95% CI = 8.0-9.3 pp] for lipid-lowering medications) and when adherent in the initial quarter (14.4 pp [13.8-15.1 pp] for lipid-lowering medications). Adherence in the initial and previous quarter increased predicted adherence considerably (22.7 pp [22.1-23.3 pp]). Similar patterns held for the antihypertensive medication cohort (antihypertensive medications) and the oral antidiabetes medication cohort (oral antidiabetes medications). The area under the curve (AUC) showed considerable improvement when moving from pooled probit models to dynamic random-effects probit models. AUC for the dynamic models exceeded 0.85 in the 3 medication cohorts, whereas the pooled probit models remained under 0.7. CONCLUSIONS: Adherence in the previous quarter is associated with adherence in the current quarter, after accounting for sources of observable and unobservable heterogeneity across enrollees. In addition, the initial value of adherence matters when explaining the likelihood of adherence.

Understanding no-show behaviour for cervical cancer screening appointments among hard-to-reach women in Bogotá, Colombia: A mixed-methods approach

The global burden of cervical cancer remains a concern and higher early mortality rates are associated with poverty and limited health education. However, screening programs continue to face implementation challenges, especially in developing country contexts. In this study, we use a mixed-methods approach to understand the reasons for no-show behaviour for cervical cancer screening appointments among hard-to-reach low-income women in Bogotá, Colombia. In the quantitative phase, individual attendance probabilities are predicted using administrative records from an outreach program (N = 23384) using both LASSO regression and Random Forest methods. In the qualitative phase, semi-structured interviews are analysed to understand patient perspectives (N = 60). Both inductive and deductive coding are used to identify first-order categories and content analysis is facilitated using the Framework method. Quantitative analysis shows that younger patients and those living in zones of poverty are more likely to miss their appointments. Likewise, appointments scheduled on Saturdays, during the school vacation periods or with lead times longer than 10 days have higher no-show risk. Qualitative data shows that patients find it hard to navigate the service delivery process, face barriers accessing the health system and hold negative beliefs about cervical cytology.

Comparing and validating medication complexity from insurance claims against electronic health records

BACKGROUND: Patient effort to comply with complex medication instructions is known to be related to nonadherence and subsequent medical complications or health care costs. A widely used Medication Regimen Complexity Index (MRCI) has been used with electronic health records (EHRs) to identify patients who could benefit from pharmacist intervention. A similar claims-derived measure may be better suited for clinical decision support, since claims offer a more complete view of patient care and health utilization. OBJECTIVE: To define and validate a novel insurance claims-based medication complexity score (MCS) patterned after the widely used MRCI, derived from EHRs. METHODS: Insurance claims and EHR data were provided by HealthPartners (N = 54,988) (Bloomington, Minnesota) and The Johns Hopkins Health System (N = 28,589) (Baltimore, Maryland) for years 2013 and 2017, respectively. Yearly measures of medication complexity were developed for each patient and evaluated with one another using rank correlation within different clinical subgroupings. Indicators for the presence of individually complex prescriptions were also developed and assessed using exact agreement. Complexity measures were then correlated with select covariates to further validate the concordance between MCS and MRCI with respect to clinical metrics. These included demographic, comorbidity, and health care utilization markers. Prescribed medications in each system's EHR were coded using the previously validated MRCI weighting rules. Insurance claims for retail pharmacy medications were coded using our novel MCS, which closely followed MRCI scoring rules. RESULTS: EHR-based MRCI and claims-based MCS were significantly correlated with one another for most clinical subgroupings. Likewise, both measures were correlated with several covariates, including count of active medications and chronic conditions. The MCS was, in most cases, more associated with key health covariates than was MRCI, although both were consistently significant. We found that the highest correlation between MCS and MRCI is obtained with patients who have similar counts of pharmacy records between EHRs and claims (HealthPartners: P = 0.796; Johns Hopkins Health System: P = 0.779). CONCLUSIONS: The findings suggest good correspondence between MCS and MRCI and that claims data represent a useful resource for assessing medication complexity. Claims data also have major practical advantages, such as interoperability across health care systems, although they lack the detailed clinical context of EHRs. DISCLOSURES: The Johns Hopkins University holds the copyright to the Adjusted Clinical Groups (ACG) system and receives royalties from the global distribution of the ACG system. This revenue supports a portion of the authors' salary. No additional or external funding supported this work. The authors have no conflict of interest to disclose.

Prior Cardiovascular Treatments-A Key Characteristic in Determining Medication Adherence After an Acute Myocardial Infarction

Objective: To investigate long-term adherence to guideline-recommended cardioprotective medications following hospitalization for an acute myocardial infarction (AMI), and identify characteristics associated with adherence.

Methods: An Australian population-based cohort study was used to identify participants who had their first AMI between 2006 and 2014 and were alive after 12 months. Linked routinely collected hospital, and prescription medication claims data was used to study adherence over time. Predictors and rates of adherence to both lipid-lowering medication and renin-angiotensin system blockade at 12 months post-AMI was assessed.

Results: 14,200 people (mean age 69.9 years, 38.7% female) were included in our analysis. At 12 months post-AMI, 29.5% (95% CI: 28.8–30.3%) of people were adherent to both classes of medication. Individuals receiving treatment with both lipid-lowering medication and renin-angiotensin system blockade during the 6 months prior to their AMI were over 9 times more likely to be adherent to both medications at 12 months post-AMI (66.2% 95% CI: 64.8–67.5%) compared to those with no prior medication use (treatment naïve) (7.1%, 95% CI: 6.4–7.9%). Prior cardiovascular treatment was the strongest predictor of long-term adherence even after adjusting for age, sex, education and income.

Conclusions: Despite efforts to improve long-term medication adherence in patients who have experienced an acute coronary event, considerable gaps remain. Of particular concern are people who are commencing guideline-recommended cardioprotective medication at the time of their AMI. The relationship between prior cardiovascular treatments and post AMI adherence offers insight into the support needs for the patient. Health care intervention strategies, strengthened by enabling policies, are needed to provide support to patients through the initial months following their AMI.

Patient Adherence to Therapy After Myocardial Infarction: A Scoping Review

BACKGROUND

Patients with myocardial infarction have low adherence to secondary prevention. Patients with acute coronary syndromes usually decide not to take cardiac drugs for 7 days after discharge for various reasons and adherence rates are usually very low. The aim of this scoping review was to identify factors influencing treatment adherence after myocardial infarction and the role of interventions to improve treatment adherence.

METHODS

Two electronic databases (PubMed and Web of Science) were systematically searched for relevant published reviews of interventions for adherence after myocardial infarction. Inclusion criteria were study design: randomized control trial, systematic reviews; published in English; sample age ≥18 years. The methodological framework proposed by Arksey & O'Malley was used to guide the review process of the study.

RESULTS

Thirteen articles met the inclusion/exclusion criteria. Four of the thirteen studies assessed factors influencing patient adherence to therapy after myocardial infarction, the remaining studies examined various interventions increasing adherence to treatment after myocardial infarction.

CONCLUSION

There is a need to improve adherence of patients to treatment after myocardial infarction. Studies show that the use of modern technologies and communication with the patients by phone improve adherence to treatment.

Machine Learning to Predict Tamoxifen Nonadherence Among US Commercially Insured Patients With Metastatic Breast Cancer

PURPOSE

Adherence to tamoxifen citrate among women diagnosed with metastatic breast cancer can improve survival and minimize recurrence. This study aimed to use real-world data and machine learning (ML) methods to classify tamoxifen nonadherence.

METHODS

A cohort of women diagnosed with metastatic breast cancer from 2012 to 2017 were identified from IBM MarketScan Commercial Claims and Encounters and Medicare claims databases. Patients with < 80% proportion of days coverage in the year following treatment initiation were classified as nonadherent. Training and internal validation cohorts were randomly generated (4:1 ratio). Clinical procedures, comorbidity, treatment, and health care encounter features in the year before tamoxifen initiation were used to train logistic regression, boosted logistic regression, random forest, and feedforward neural network models and were internally validated on the basis of area under receiver operating characteristic curve. The most predictive ML approach was evaluated to assess feature importance.

RESULTS

A total of 3,022 patients were included with 40% classified as nonadherent. All models had moderate predictive accuracy. Logistic regression (area under receiver operating characteristic 0.64) was interpreted with 94% sensitivity (95% CI, 89 to 92) and 0.31 specificity (95% CI, 29 to 33). The model accurately classified adherence (negative predictive value 89%) but was nondiscriminate for nonadherence (positive predictive value 48%). Variable importance identified top predictive factors, including age ≥ 55 years and pretreatment procedures (lymphatic nuclear medicine, radiation oncology, and arterial surgery).

CONCLUSION

ML using baseline administrative data predicts tamoxifen nonadherence. Screening at treatment initiation may support personalized care, improve health outcomes, and minimize cost. Baseline claims may not be sufficient to discriminate adherence. Further validation with enriched longitudinal data may improve model performance.

Statistical methods versus machine learning techniques for donor-recipient matching in liver transplantation

Donor-Recipient (D-R) matching is one of the main challenges to be fulfilled nowadays. Due to the increasing number of recipients and the small amount of donors in liver transplantation, the allocation method is crucial. In this paper, to establish a fair comparison, the United Network for Organ Sharing database was used with 4 different end-points (3 months, and 1, 2 and 5 years), with a total of 39, 189 D-R pairs and 28 donor and recipient variables. Modelling techniques were divided into two groups: 1) classical statistical methods, including Logistic Regression (LR) and Naïve Bayes (NB), and 2) standard machine learning techniques, including Multilayer Perceptron (MLP), Random Forest (RF), Gradient Boosting (GB) or Support Vector Machines (SVM), among others. The methods were compared with standard scores, MELD, SOFT and BAR. For the 5-years end-point, LR (AUC = 0.654) outperformed several machine learning techniques, such as MLP (AUC = 0.599), GB (AUC = 0.600), SVM (AUC = 0.624) or RF (AUC = 0.644), among others. Moreover, LR also outperformed standard scores. The same pattern was reproduced for the others 3 end-points. Complex machine learning methods were not able to improve the performance of liver allocation, probably due to the implicit limitations associated to the collection process of the database.

Predicting adjuvant endocrine therapy initiation and adherence among older women with early-stage breast cancer

PURPOSE

The CALGB 9343 trial demonstrated that women age 70 or older with early-stage, estrogen receptor positive (ER +) breast cancer (BC) may safely forgo radiation therapy (RT) and be treated with breast conserving surgery followed by adjuvant endocrine therapy (AET) alone. However, most patients in this population still undergo RT in part because AET adherence is low. We sought to develop a predictive model for AET initiation and adherence in order to improve decision-making with respect to RT omission.

METHODS

Women ages 70 and older with early-stage, ER + BC were identified using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database. Comorbidities, socioeconomic measures, prescription medications, and demographics were collected as potential predictors. Bivariate analysis was performed to identify factors associated with AET initiation and adherence. Stepwise selection of significant predictors was used to develop logistic regression classifiers for initiation and adherence. Model performance was evaluated using the c-statistic and other measures.

RESULTS

11,037 patients met inclusion criteria. Within the cohort, 8703 (78.9%) patients initiated AET and 6685 (60.6%) were adherent to AET over 1 year. Bivariate predictors of AET initiation were similar to predictors of adherence. The best AET initiation and adherence classifiers were poorly predictive with c-statistics of 0.65 and 0.60, respectively.

CONCLUSIONS

The best models in the present study were poorly predictive, demonstrating that the reasons for initiation and adherence to AET are complex and individual to the patient, and therefore difficult to predict. Initiation and adherence to AET are important factors in decision-making regarding whether or not to forgo adjuvant RT. In order to better formulate treatment plans for this population, future work should focus on improving individual prediction of AET initiation and adherence.

Novel application of approaches to predicting medication adherence using medical claims data

OBJECTIVE

To compare predictive analytic approaches to characterize medication nonadherence and determine under which circumstances each method may be best applied.

DATA SOURCES/STUDY SETTING

Medicare Parts A, B, and D claims from 2007 to 2013.

STUDY DESIGN

We evaluated three statistical techniques to predict statin adherence (proportion of days covered [PDC ≥ 80 percent]) in the year following discharge: standard logistic regression with backward selection of covariates, least absolute shrinkage and selection operator (LASSO), and random forest. We used the C-index to assess model discrimination and decile plots comparing predicted values to observed event rates to evaluate model performance.

DATA EXTRACTION

We identified 11 969 beneficiaries with an acute myocardial infarction (MI)-related admission from 2007 to 2012, who filled a statin prescription at, or shortly after, discharge.

PRINCIPAL FINDINGS

In all models, prior statin use was the most important predictor of future adherence (OR = 3.65, 95% CI: 3.34-3.98; OR = 3.55). Although the LASSO regression model selected nearly 90 percent of all candidate predictors, all three analytic approaches had moderate discrimination (C-index ranging from 0.664 to 0.673).

CONCLUSIONS

Although none of the models emerged as clearly superior, predictive analytics could proactively determine which patients are at risk of nonadherence, thus allowing for timely engagement in adherence-improving interventions.