Inductive thematic analysis of healthcare qualitative interviews using open-source large language models: How does it compare to traditional methods?

BACKGROUND: Large language models (LLMs) are generative artificial intelligence that have ignited much interest and discussion about their utility in clinical and research settings. Despite this interest there is sparse analysis of their use in qualitative thematic analysis comparing their current ability to that of human coding and analysis. In addition, there has been no published analysis of their use in real-world, protected health information. OBJECTIVE: Here we fill that gap in the literature by comparing an LLM to standard human thematic analysis in real-world, semi-structured interviews of both patients and clinicians within a psychiatric setting. METHODS: Using a 70 billion parameter open-source LLM running on local hardware and advanced prompt engineering techniques, we produced themes that summarized a full corpus of interviews in minutes. Subsequently we used three different evaluation methods for quantifying similarity between themes produced by the LLM and those produced by humans. RESULTS: These revealed similarities ranging from moderate to substantial (Jaccard similarity coefficients 0.44-0.69), which are promising preliminary results. CONCLUSION: Our study demonstrates that open-source LLMs can effectively generate robust themes from qualitative data, achieving substantial similarity to human-generated themes. The validation of LLMs in thematic analysis, coupled with evaluation methodologies, highlights their potential to enhance and democratize qualitative research across diverse fields.

TimelinePTC: Development of a unified interface for pathways to care collection, visualization, and collaboration in first episode psychosis.

This paper presents TimelinePTC, a web-based tool developed to improve the collection and analysis of Pathways to Care (PTC) data in first episode psychosis (FEP) research. Accurately measuring the duration of untreated psychosis (DUP) is essential for effective FEP treatment, requiring detailed understanding of the patient's journey to care. However, traditional PTC data collection methods, mainly manual and paper-based, are time-consuming and often fail to capture the full complexity of care pathways. TimelinePTC addresses these limitations by providing a digital platform for collaborative, real-time data entry and visualization, thereby enhancing data accuracy and collection efficiency. Initially created for the Specialized Treatment Early in Psychosis (STEP) program in New Haven, Connecticut, its design allows for straightforward adaptation to other healthcare contexts, facilitated by its open-source codebase. The tool significantly simplifies the data collection process, making it more efficient and user-friendly. It automates the conversion of collected data into a format ready for analysis, reducing manual transcription errors and saving time. By enabling more detailed and consistent data collection, TimelinePTC has the potential to improve healthcare access research, supporting the development of targeted interventions to reduce DUP and improve patient outcomes.

Racial and ethnic inequities in spatial access to pharmacies: A geographic information system analysis.

BACKGROUND: Pharmacy accessibility is crucial for equity in health care access because community pharmacists may reach individuals who do not have access to other health care providers. OBJECTIVE: The objective of this study was to determine whether spatial access to pharmacies differs among racial/ethnic groups across the rural-urban continuum. METHODS: We obtained a 30% random sample of the Research Triangle Institute synthetic population, sampled at the census block level. For each individual, we defined optimal pharmacy access as having a driving distance ≤2 miles to the closest pharmacy in urban counties, ≤5 miles in suburban counties, and ≤10 miles in rural counties. We used a logistic regression model to measure the association between race/ethnicity and pharmacy access, while controlling for racial/ethnic composition of the census tract, area deprivation index, income, age, gender, and U.S. region. The model included an interaction between race/ethnicity and urbanicity to evaluate whether racial/ethnic inequities differed across the rural-urban continuum. RESULTS: The sample included 90,749,446 individuals of whom 80.6% had optimal pharmacy access. Racial/ethnic inequities in pharmacy access differed across the rural-urban continuum (P value for interaction= <0.0001). In rural areas, Black (OR 0.87; 95% CI 0.86-0.87), Hispanic (OR 0.80; 95% CI 0.79-0.80), and indigenous (OR 0.47; 95% CI 0.47-0.48) individuals had lower odds of optimal pharmacy access, than White individuals. Hispanic (OR 0.96; 95% CI 0.96-0.97) and Indigenous individuals (OR 0.75; 95% CI 0.75-0.76) had lower odds of optimal pharmacy access compared to White individuals in suburban areas. In Western states, Asian had lower odds of optimal pharmacy access in suburban (OR 0.88; 95% CI 0.86-0.90) and rural areas (OR 0.91; 95% CI 0.87-0.95) compared to White individuals. CONCLUSIONS: Racial/ethnic inequities in spatial access to community pharmacies vary between urban and rural communities. Underrepresented racial/ethnic groups have significantly lower pharmacy access in rural and some suburban areas, but not in urban areas.

Understanding Barriers and Facilitators of Primary Care Use Among Assertive Community Treatment Teams Via Qualitative Analysis of Clients and Clinicians.

Individuals with severe mental illness and substance use disorders face complex barriers to achieving physical health. This study aims to explore the barriers and facilitators of primary care access among an Assertive Community Treatment (ACT) team. Semi-structured qualitative interviews were conducted with 14 clients and 7 clinicians from an ACT team at a community mental health center in Connecticut. Data analysis followed a grounded theory approach, with codes and themes emerging iteratively during the interview process. The study identified multifaceted barriers to accessing primary care, including economic challenges, homelessness, and the prioritization of mental health and substance use symptoms over healthcare. The conceptual framework consists of nine dominant themes: clients' attitudes, knowledge, mental health, and motivations ("Client-Level Barriers and Facilitators"); ACT team-directed care coordination and relationship-building as well as primary care provider communication ("Provider-Level Barriers and Facilitators"); and clients' experiences with medical care and socioeconomic status ("Systemic-Level Barriers and Facilitators"). This research provides valuable insights into the various barriers faced by ACT clients in accessing primary care. Improving primary care access for individuals with severe mental illness and substance use disorders is crucial for reducing health disparities in this vulnerable population.

The Anomalous Effect of COVID-19 Pandemic Restrictions on the Duration of Untreated Psychosis (DUP).

We investigated the impact of COVID-19 restrictions on the duration of untreated psychosis (DUP). First-episode psychosis admissions (n=101) to STEP Clinic in Connecticut showed DUP reduction (p=.0015) in the pandemic, with the median reducing from 208 days during the pre-pandemic to 56 days in the early pandemic period and subsequently increasing to 154 days (p=.0281). Time from psychosis onset to anti-psychotic prescription decreased significantly in the pandemic (p=.0183), with the median falling from 117 to 35 days. This cohort study demonstrates an association between greater pandemic restrictions and marked DUP reduction and provides insights for future early detection efforts.

Accessibility of Pulmonary Rehabilitation in the US.

This cross-sectional study examines nationwide travel time to pulmonary rehabilitation (PR) programs and PR access in the US.

Role of independent versus chain pharmacies in providing pharmacy access: a nationwide, individual-level geographic information systems analysis.

Pharmacy accessibility is critical for equity in medication access and is jeopardized by pharmacy closures, which disproportionately affect independent pharmacies. We conducted a geographic information systems analysis to quantify how many individuals across the US do not have optimal pharmacy access or solely rely on independent pharmacies for access. We generated service areas of pharmacies using OpenStreetMap data. For each individual in a 30% random sample of the 2020 RTI US Household Synthetic Population™ (n=90,778,132), we defined optimal pharmacy access as having a driving distance to the closest pharmacy ≤2 miles in urban counties, ≤5 miles in suburban counties, and ≤10 miles in rural counties. Individuals were then categorized according to their access to chain and independent pharmacies. Five percent of the sample or ~15.1 million individuals nationwide relied on independent pharmacies for optimal access. Individuals relying on independent pharmacies for optimal access were more likely to live in rural areas, be 65 years or older, and belong to low-income households. Another 19.5% of individuals in the sample did not have optimal pharmacy access, which corresponds to 59.0 million individuals nationwide. Our findings demonstrate that independent pharmacies play a critical role in ensuring equity in pharmacy access.

Mapping Heat Vulnerability of a Community Mental Health Center Population.

Individuals with serious mental illness are vulnerable to extreme heat due to biological, social, and place-based factors. We examine the spatial correlation of prevalence of individuals treated at a community mental health center to heat vulnerability. We applied a heat vulnerability index (HVI) to the catchment of the Connecticut Mental Health Center in New Haven, Connecticut. Geocoded addresses were mapped to correlate patient prevalence with heat vulnerability of census tracts. Census tracts closer to the city center had elevated vulnerability scores. Patient prevalence was positively correlated with HVI score (Pearson's r(44) = 0.67, p < 0.01). Statistical significance persists after correction for spatial autocorrelation (modified t-test p < 0.01). The study indicates that individuals treated at this community mental health center are more likely to live in census tracts with high heat vulnerability. Heat mapping strategies can help communicate risk and target resources at the local scale.

Availability and Accessibility of Live Nonreplicating Smallpox/Mpox Vaccine.

This cross-sectional study compares the availability and accessibility of mpox vaccine sites with the number of reported cases and allocated vaccines.

Travel Times to Facilities Offering Oral Medication Therapy for COVID-19.

This quality improvement study evaluates access to oral COVID-19 therapeutics within communities in the 48 contiguous states and the District of Columbia.

Impact of a Medical Clinic Relocation on Travel Time: A Tale of 2 Modes of Transportation.

The aim of this study was to quantify access disparities by examining the impact of a medical clinic relocation on travel time differences for patients using private cars versus public transit. Longitude and latitude of patient home addresses were extracted from electronic medical records for the 4 years before the clinic move. Using offline, open-source, and HIPAA-compliant routing software, roundtrip travel times were computed from each home address to the old and new clinic locations via car and bus. Mean roundtrip travel time by a car changed from 41.3 (IQR 16.1-80.7) to 45.4 (IQR 25.9-78.1) minutes, a 9.9% increase. Mean roundtrip travel time by public transit changed from 67.5 (IQR 51.5-100.2) to 120.8 (100.3-156.1) minutes, a 78.9% increase. Even clinic relocations that minimally impact car travel times can nevertheless yield substantial changes to those traveling by public transit. Clinics and health systems that wish to reduce barriers to accessing health care, especially among those already facing structural inequities, would benefit from utilizing the analytic approach described here.

Granular analysis of pathways to care and durations of untreated psychosis: A marginal delay model.

OBJECTIVE: An extensive international literature demonstrates that understanding pathways to care (PTC) is essential for efforts to reduce community Duration of Untreated Psychosis (DUP). However, knowledge from these studies is difficult to translate to new settings. We present a novel approach to characterize and analyze PTC and demonstrate its value for the design and implementation of early detection efforts. METHODS: Type and date of every encounter, or node, along the PTC were encoded for 156 participants enrolled in the clinic for Specialized Treatment Early in Psychosis (STEP), within the context of an early detection campaign. Marginal-delay, or the portion of overall delay attributable to a specific node, was computed as the number of days between the start dates of contiguous nodes on the PTC. Sources of delay within the network of care were quantified and patient characteristic (sex, age, race, income, insurance, living, education, employment, and function) influences on such delays were analyzed via bivariate and mixed model testing. RESULTS: The period from psychosis onset to antipsychotic prescription was significantly longer (52 vs. 20.5 days, [p = 0.004]), involved more interactions (3 vs. 1 nodes, [p<0.001]), and was predominated by encounters with non-clinical nodes while the period from antipsychotic to STEP enrollment was shorter and predominated by clinical nodes. Outpatient programs were the greatest contributor of marginal delays on both before antipsychotic prescription (median [IQR] of 36.5 [1.3-132.8] days) and (median [IQR] of 56 [15-210.5] days). Sharper functional declines in the year before enrollment correlated significantly with longer DUP (p<0.001), while those with higher functioning moved significantly faster through nodes (p<0.001). No other associations were found with patient characteristics and PTCs. CONCLUSIONS: The conceptual model and analytic approach outlined in this study give first episode services tools to measure, analyze, and inform strategies to reduce untreated psychosis.

Navigating the digital divide: providing services to people with serious mental illness in a community setting during COVID-19.

Community mental healthcare around the world has been strained as people need more help and experience more barriers to access due to COVID-19. The rapid shift to telehealth services necessitated by the pandemic has made these difficulties even more pronounced. While this transition presented challenges for nearly every healthcare system, it has proven especially difficult for low resource settings such as community health centers. This article is a critical observational study of the care transformation of a state-funded safety net psychiatric system responding to the clinical needs of patients during the COVID-19 pandemic. By discussing the challenges, opportunities, and creative solutions for staff and patients, the article highlights the new importance of technology and adaptability in clinical care and outlines clear recommendations to ensure vulnerable populations do not fall into the "digital divide."

Pharmacy deserts: More than where pharmacies are.

BACKGROUND: In the United States, geographic access is a major driver of health care disparities. Studies have shown that pharmacy deserts are prevalent in the United States, even in major metropolitan areas. However, one limitation often cited by these studies is the use of distance rather than travel time to define pharmacy deserts. OBJECTIVE: The aim of this study was to assess pharmacy deserts using travel time and to provide a more holistic approach by incorporating analysis of private vehicles and public transportation. METHODS: Pharmacy details were collected from the National Provider Identifier database and neighborhood characteristics from collected census data for the four largest U.S. cities. Pharmacy access was evaluated using open-source routing engines. We determined neighborhoods in pharmacy deserts using both distance and travel time analyses. Sensitivity analysis was performed to determine changes to pharmacy deserts based on small changes in travel time. RESULTS: Of 4654 neighborhoods identified in the four cities of interest, 670 (14.4%) neighborhoods were in pharmacy deserts based on distance. Despite accounting for 28.9% of all neighborhoods, predominantly white neighborhoods only accounted for 4.3% of pharmacy deserts. When evaluating pharmacy deserts by car and public transportation, predominantly white neighborhoods accounted for 2.3% and 1.7% of total pharmacy deserts, respectively. Finally, by reducing travel time from 15 minutes to 10 minutes, pharmacy deserts by car and public transportation increased by 105% and 199%, respectively. All but 3 of the new pharmacy deserts found in the sensitivity analysis were found in nonpredominantly white neighborhoods. CONCLUSION: Using travel time and incorporating modes of transportation, we found that disparities in pharmacy access are more than just where pharmacies are located geographically. There are additional layers of disparities, such as access to public transportation, that need to be addressed to reduce the number of pharmacy deserts.

Reducing the Duration of Untreated Psychosis (DUP) in a US Community: A Quasi-Experimental Trial.

Objective: Duration of Untreated Psychosis (DUP) remains unacceptably long and limits effectiveness of care. To determine whether an early detection campaign ("Mindmap") can reduce DUP in a US community setting. Methods: In this nonrandomized controlled trial, Mindmap targeted the catchment of one specialty first-episode service or FES (STEP, Greater New Haven) from 2015 to 2019, while usual detection efforts continued at a control FES (PREP, Greater Boston). Mindmap targeted diverse sources of delay through mass & social media messaging, professional outreach & detailing, and rapid enrollment of referrals. Both FES recruited 16-35 years old with psychosis onset ≤3 years. Outcome measures included DUP-Total (onset of psychosis to FES enrollment), DUP-Demand (onset of psychosis to first antipsychotic medication), and DUP-Supply (first antipsychotic medication to FES enrollment). Results: 171 subjects were recruited at STEP and 75 at PREP. Mindmap was associated with an increase in the number of referrals and in efficiency of engagement at STEP. Pre-campaign DUP (2014-2015) was equivalent, while Mindmap was associated with DUP reductions at STEP but not PREP. DUP-Total fell significantly in both the first and the second quartile (11.5 and 58.5 days reduction per campaign year, respectively). DUP-Demand and DUP-Supply fell in the third quartiles only (46.3 and 70.3 days reduction per campaign year, respectively). No reductions were detectable across all quartiles at PREP, but between site comparisons were not significant. Conclusions: This is the first controlled demonstration of community DUP reduction in the US, and can inform future early detection efforts across diverse settings.

Taking the next step: Improving care transitions from a first-episode psychosis service.

AIMS: First-episode services (FES) improve outcomes in recent onset psychosis, but there is growing concern about how patients fare after discharge from these time-limited services. METHODS: A quality improvement approach (QI) was used to improve patient engagement in the discharge planning process (disposition), and successful engagement in care 3 months after discharge from the FES (transfer). Data from 144 consecutive discharges over 62 months are presented. A planning phase was followed by recurrent Plan-Do-Study-Act cycles (PDSA) that included the introduction of proactive efforts targeting disposition planning (with patients and families) and follow-up to facilitate transfer after discharge. Fisher's exact test was used to compare disposition and transfer outcomes across the QI phases. RESULTS: This QI approach was sustained through a three-fold escalation in discharge volume. Transfer status at 3 months was significantly different between the pre- and post PDSA phases (p = .02). A greater proportion were confirmed transfers post-PDSA (54.3 vs. 37%), but of those with known status at 3 months, similar proportions were successfully transferred (76, 73%). Patients discharged post-PDSA were less likely to have unknown treatment status (26 vs. 51%). Disposition outcomes were also significantly improved post-PDSA (p = .03). Patients were more likely to engage with discharge planning (69.7 vs. 48.6%) and less likely to be lost to follow-up (13.8 vs. 25.7%), or to refuse assistance (11.0 vs. 20.0%). CONCLUSION: This QI approach offers a feasible way to improve disposition and transfer after FES and can be built upon in efforts to sustain functional gains in onward pathways.

Corrigendum to: Reducing the Duration of Untreated Psychosis (DUP) in a US Community: A Quasi-Experimental Trial.

[This corrects the article DOI: 10.1093/schizbullopen/sgab057.].

First help-seeking attempt before and after psychosis onset: measures of delay and aversive pathways to care.

PURPOSE: Delay in receiving effective treatment for psychosis adversely impacts outcomes. We investigated the timing of the first help-seeking attempt in individuals with recent onset non-affective psychosis by comparing those who sought help during the prodrome to those who sought help after psychosis onset across sociodemographic and clinical characteristics, overall functioning, and occurrence of aversive events during their pathways to care. METHODS: Patients were admitted from February 1st, 2014 to January 31st, 2019 to the Program for Specialized Treatment Early in Psychosis (STEP) in New Haven, CT. Psychosis-onset date was ascertained using the Structured Interview for Psychosis-risk Syndromes. Key dates before and after psychosis onset, along with initiators and aversive events, were collected via semi-structured interview. RESULTS: Within 168 individuals, 82% had their first help-seeking episode after psychosis onset and did not differ in terms of sociodemographic characteristics from prodrome help seekers. When the first help-seeking episode started before (i.e., during prodrome) vs after psychosis onset it was mostly initiated by patients vs family members (Cramer's V = 0.23, p = 0.031) and led to a faster prescription of an antipsychotic once full-blown psychosis emerged (time to antipsychotic since psychosis onset = 21 vs 56 days, p = 0.03). No difference in aversive events before STEP enrollment was detected across groups. CONCLUSION: Help seeking during the prodrome is associated with faster initiation of antipsychotic treatment and is more likely to be self-initiated, compared to help seeking after psychosis onset. Early detection efforts that target prodromal samples may improve the length and experience of pathways to care.