Professional Abstract
PCORI funded the Pilot Projects to explore how to conduct and use patient-centered outcomes research in ways that can better serve patients and the healthcare community. Learn more.
Background
Depression, a recurrent, chronic, mental health condition, affects more than 350 million people worldwide. Depression is undiagnosed, untreated, and/or inadequately treated among the majority of those with the condition.
Understanding how patients engage in depression management decisions may help facilitate patient–provider and patient–family communication regarding symptom recognition and management; identify patient, family, and provider information needs; better match individual preferences and treatment needs; improve treatment adherence; and optimize treatment response. Health information technology is an important tool for patients, providers, and healthcare systems to identify, manage, and monitor chronic conditions like depression.
Project Purpose
The researchers developed, tested, and refined a stakeholder-centered decision-support tool for depression within a tribally owned and operated healthcare organization.
Study Design
The researchers used qualitative and quantitative methods to develop, test, and refine a stakeholder centered decision-support tool to improve communication and better incorporate the perspectives of Southcentral Foundation (SCF)'s stakeholders (e.g., customer-owners, providers, and leaders) into healthcare decisions. The three specific aims of this mixed-methods, prospective cohort study were as follows:
- Specific Aim 1: Identify stakeholder understanding, preferences, and needs that influence depression treatment decisions in Alaska Native and American Indian (ANAI) individuals.
- Specific Aim 2: Develop, pilot, and evaluate a stakeholder-centered decision-support tool to help translate and integrate evidence-based guidelines and stakeholder understanding, preferences, and needs into depression management decisions.
- Specific Aim 3: Determine the impact of the stakeholder-centered decision-support tool on health, service utilization, and economic outcomes.
Participants, Interventions, Settings, and Outcomes
The study was conducted in three stages. An iterative search process commonly used in community-based participatory research and community engagement projects was used to identify stakeholder understanding, preferences, and needs in Specific Aim 1. The researchers used a purposive sampling strategy to recruit stakeholders for Aim 1(n = 38). They conducted semi-structured interviews with these stakeholders and used a thematic network analysis to identify common views across stakeholders. These views were presented to a steering committee for guidance on developing the decision support tool.
In Specific Aim 2, tool feasibility and acceptability was assessed with 20 customer-owners diagnosed with depression.
In Specific Aim 3, the decision-support tool was deployed to 131 ANAI adults who screened positive for depression in three randomly selected primary care clinics. Of these, three withdrew after the iPad-based shared decision-support tool crashed, one withdrew because of fatigue, one did not consent to medical record review, one was disenrolled due to ineligibility, and 13 were excluded because they were seen in both the control and intervention clinic.
The control group consisted of 263 ANAI adults who screened positive for depression in the treatment-as-usual condition during the same time period.
Data Analysis
Demographic and health factors included age, gender, marital status, insurance status (as a proxy for socioeconomic status), median income of residential ZIP Code, number of chronic physical conditions (e.g., cardiovascular disease, cancer, diabetes, epilepsy, obesity), and prior service utilization counts. A random effect intercept for the primary care clinic team accounted for correlation due to nesting and allowed for adjustments according to provider factors including primary care provider (PCP) age, PCP gender, PCP tenure with SCF, and PCP position type (physician, interim provider, midlevel practitioner (e.g., nurse practitioner, physician assistant), senior physician, or medical director). Researchers used multilevel models including: Poisson regression for count outcomes; zero-inflated Poisson regression for count outcomes with a high number of zerocounts; log-transformed linear regression; and Cox proportional hazards regression for time-to-event data.
Findings
Although few differences in univariable analyses were found, when demographics, healthcare system service utilization in the pre-period, diagnoses in the pre-period, and provider factors were accounted for as covariates in multilevel models, significant differences in healthcare system service utilization in the six month post-period were observed. Use of the tool decreased primary care utilization (RR 95% Cl: 0.51-0.88) and behavioral health consultant visits (RR 95% Cl: 0.13-0.52). Use of the tool did not affect emergency department (ED)/urgent care (UC) utilization (RR 95% Cl: 0.35-1.34), the number of depression medication dispenses (RR 95% Cl: 0.74-1.68), days-supply of depression medication (βlog 95% Cl: -2.73-2.76), or days to first depression medication refill for those previously dispensed medications (n = 186; RR 95% Cl: 0.46-1.71). The number of times screened with the Patient Health Questionnaire (PHQ)-9 in the follow-up period was significantly higher for those who utilized the tool in univariable analysis, but not statistically significant when adjusted for demographic and health factors (RR 95% Cl: 0.98-2.40). Healthcare service costs related to utilization, specifically primary care, behavioral health, and ED/UC charges, did not appear to be affected by tool administration. However, the six-month study time period may be too short to detect a difference, especially for ED/UC charges (p = 0.08).
Limitations
The evaluation period was shorter than anticipated because of delays in tool development. Providers often did not readminister the PHQ-9 at follow-up appointments to guide depression management. Moreover, documentation for those who did return to primary care for follow‑up was often limited to medical notes with limited search ability. Patients who felt better and did not return to the clinic for whatever reason were lost to follow up. Economic evaluation is based on billing data and may not represent the true cost to SCF's healthcare system.
Conclusions
Patient engagement and tailored electronic depression management resources may help guide healthcare management and resource allocation decisions. Although few statistically significant changes in patient health outcomes or healthcare system charges were found during the tool pilot period, observed changes in healthcare service utilization suggest that the tool may influence depression outcomes, service utilization, and healthcare costs.