De-Implementing Opioids for Dental Extractions (DIODE): a multi-clinic, cluster-randomized trial of clinical decision support strategies in dentistry

Design

The study was a prospective, cluster-randomized trial of CDS to de-implement opioid prescribing. Due to the impracticality of randomizing at the patient level, randomization occurred at the provider level, whereby dental providers were randomly assigned on a 1:1:1 basis to one of three study arms: standard practice (SP), clinical decision support (CDS), or clinical decision support augmented with patient educational materials (CDS-E). Randomization was stratified by provider type (dentist vs. oral surgeon) and providers’ volume of opioid prescribing in the baseline period. The study statistician (SA) generated the randomization sequence. Fifty providers were randomized, with 1 subsequently excluded due to a very low volume of extractions during the intervention period. The analysis sample consisted of 49 providers with 12,924 unique patients. Our hypothesis was that the CDS and CDS-E arms would show greater reductions in opioid prescribing compared to standard practice. The outcome of whether or not an opioid was prescribed was assessed at the patient level, and patients were blind to their provider’s study arm. Figure 1 depicts the study flow. The study protocol and methodological details have been previously reported [19]. The HealthPartners Institutional Review Board approved the study.

Fig. 1figure 1

Study flow

CDS, clinical decision support; CDS-E, clinical decision support with patient education; SP, standard practice

Setting

This study was conducted at 22 HealthPartners dental clinics in Minnesota. HealthPartners is the largest consumer-governed non-profit healthcare organization in the USA. It was an early adopter of health information technology and all dental clinics use an electronic health record (EHR) that integrates dental and medical records. Many providers practiced across multiple clinics during the study period, with only 6 of 49 providers completing all of their extraction encounters at a single clinic.

Eligibility

The CDS was considered a quality improvement initiative of the HealthPartners Dental Group, and the project had support and approval from leadership within the organization. The local IRB considered this approval as an alternative to individual, written informed consent for dentists enrolled in the study. Thus, all eligible dentists and oral surgeons at HealthPartners were included. Dentists and oral surgeons at HealthPartners who performed at least 6 dental extractions in the year prior to launching the trial were eligible for inclusion in the study. Patients were included in the study if they were age 16 or older and had a permanent tooth extraction that was performed by an eligible provider at HealthPartners. HealthPartners routinely conducts data-driven quality improvement initiatives and research on services and patient outcomes. All patients at HealthPartners are given the opportunity to opt out of having their data used for research. Patients who had opted out of research still saw their provider as they normally would, but their data were not used in the current study.

Study arms

Dentists and oral surgeons were given access to the CDS according to the arm to which they were randomized. All providers at HealthPartners were made aware of the project, and there was no blinding. Providers in the SP arm were not given access to the CDS module, whereas the CDS was accessible for providers in the CDS and CDS-E arms. Patients were thus exposed to the intervention to which their provider was randomized. Providers in the CDS and CDS-E arms were informed about their assignment and trained on the purpose of the CDS, what health information it incorporated, and how to access the CDS through written communication with an opportunity for clarification.

Standard practice (SP)

Providers in the SP arm did not have access to the CDS in the EHR. The SP arm represents the control condition in the study.

Clinical decision support (CDS)

Providers in the CDS arm were given access to the CDS at the point-of-care through the EHR, which was accessible to providers who were seeing patients for a dental extraction, with a highlighted link in the EHR when a patient had a treatment plan involving an extraction. The CDS content was developed by the research team and content experts to provide guidance on pain management recommendations for dental extractions. Prior to deploying the CDS, the study team conducted observations of 15 dental extractions at HealthPartners to determine provider workflow related to analgesic decision-making, including review of the EHR and discussions with the patient, in order to optimize the placement and timing of the CDS within the EHR. The CDS was personalized for the patient to display potential medication interactions with commonly recommended analgesics as identified in Micromedex [20], flagged relevant health conditions that may have a bearing on pain management decisions and risk for opioid misuse (including history of substance use disorders, history of overdose, current naloxone prescription), and provided automated access to the state’s prescription drug monitoring program (PDMP). The CDS then provided guidance regarding analgesia options to consider recommending based on the available evidence with the overarching goal of providing personalized healthcare. The CDS was designed to simplify decision-making by synthesizing this relevant information in a single interface within the EHR.

Clinical decision support with patient education (CDS-E)

Providers in the CDS-E arm had the same access to the CDS as providers in the CDS arm. In addition, the front desk printed an educational handout at check-in for patients with a treatment planned extraction. This handout included information comparing opioid and non-opioid options for analgesia on risks, benefits, and effectiveness. The intention of the handout was to equip patients with information about medication options and prompt a discussion with the provider about needs, goals, and preferences. Although this type of enhanced patient-provider communication may not be very common in dentistry in the USA, it was thought that it could lead to more patient-centered care and improve patients’ acceptance of their analgesia plan. Furthermore, although patient information and decision aids are sometimes used in dentistry, examination of the existing materials at HealthPartners found that they did not provide information comparing the efficacy, safety, and side effects of the various analgesics. The handout sought to normalize the experience of some limited discomfort, set expectations, and provide tips for managing post-extraction pain (e.g., ice) and guidance for when to contact providers if they needed additional help in managing their post-extraction pain.

Study endpoint and measures

Data were derived from the EHR for each extraction encounter. The primary outcome of interest was whether an opioid was prescribed at the extraction encounter. Patient and provider characteristics were also obtained from the EHR, as was provider utilization of the CDS for each extraction encounter.

The study period spanned January 2019 through May 2021, which included a 12-month baseline (pre-intervention) period and a 15-month intervention period. The intervention period was originally planned to be 12 months, but was extended in response to the disruptions in services and delays caused by the COVID-19 pandemic. The analysis sample consisted of patients ages 16 and older seen by 49 participating providers for a dental extraction. The patient’s first extraction encounter within the study period was used as the analytical record, for a total of N = 12,924 patients (N = 7275 unique patients in the baseline period, N = 5649 unique patients in the intervention period).

Statistical analysis

A priori power analyses were based on 2018 data from HealthPartners and estimated 6900 patients ages 16 and older receiving extractions in each of the baseline and intervention periods seen by 51 providers (slightly more than the final analysis sample). In 2018, 40% of tooth extraction encounters included an opioid prescription, and the provider-level intraclass correlation for opioid prescribing was estimated at 0.3. Under these assumptions about HealthPartners providers and patients, and α = 0.05, the study had 80% power to detect a differential change of 23% from baseline to intervention periods when comparing CDS or CDS-E to SP arms.

Data were analyzed using generalized linear mixed models for binary data (binomial family with logit link function; alternatively known as mixed effects logistic regression). The dependent variable was extraction-level opioid prescription (no vs. yes). Explanatory variables include Time (baseline vs. intervention), Arm (CDS vs. CDS-E vs. SP), and the Arm × Time interaction (representing the effect of interest). Analyses were adjusted for patient sex and age, extraction complexity (simple vs. complex), and baseline prescribing strata (dentist vs. oral surgeon; baseline opioid prescribing volume). Models included a random intercept for provider to account for the cluster design. Comparisons were made from the baseline to the intervention period within each study arm, while the primary effect of interest was examined as the differential change between CDS and CDS-E arms in comparison to the SP arm. Contrasts of interest were examined using odds ratios, model-based predicted probabilities, and their confidence intervals (95% confidence intervals for baseline vs. intervention period comparisons within study arm; 97.5% confidence intervals for comparing each of the two intervention arms against SP with respect to differential change. The 97.5% confidence interval was used for the primary endpoint in order to address multiple comparisons across three study arms). Analyses were conducted on an intent-to-treat basis. To give additional context for the findings, utilization of the CDS was examined descriptively.

In addition to the primary analysis, post hoc analyses were conducted excluding oral surgeons (due to their having more opioid prescriptions and lower utilization of the CDS). The association between providers’ CDS utilization and opioid prescribing at the extraction visit was examined among providers in the CDS and CDS-E arms (i.e., those who had the opportunity to use the CDS). These post hoc analyses were conducted using a similar analytical strategy as used for the primary analysis, except that it excluded oral surgeons due to their minimal utilization of the CDS and included an interaction term between extraction complexity and CDS utilization.

留言 (0)

沒有登入
gif