Applied Clinical Informatics最新文献

Background:  Health Information Technology is increasingly being used to help providers connect patients with community resources to meet health-related social needs (e.g., food, housing, transportation). Research is needed to design efficient, simple, and engaging interfaces during a sensitive process that involves multiple stakeholders. Research is also needed to understand the roles, expectations, barriers, and facilitators these different stakeholders (i.e., patients, providers, and community-based organizations [CBOs]) face during this process.

Objectives:  We applied the human-centered design approach to develop a multi-interface social care referral platform. This approach allowed us to understand the needs of each stakeholder and address potential workflow concerns.

Methods:  This paper reports on the research team's understanding of the design process from 48 different user tests. We conducted three rounds of user testing on an interactive prototype(s) and adapted the prototype after each round.

Results:  Our results summarize several key findings useful for patients, clinical teams, and staff of CBOs when designing a social care referral platform. Our user testing highlighted that patient-facing interfaces offer tremendous opportunities to allow patients to be the leader of the social care referral process. CBOs have varying needs that must be addressed, and providing CBO staff with opportunities to connect with patients is critical. Finally, health care teams have more structured workflows. Integration within the electronic health record system provides opportunities for health care staff to support their patients more easily given these barriers.

Conclusion:  Our resulting, patient-centered platform allows patients to self-screen and self-refer to organizations that match their unmet needs.

Background: The proliferation of electronic health record (EHR) alerts has led to widespread alert fatigue and clinician burnout, undermining the effectiveness of clinical decision support and compromising patient safety.

Objective: We introduce a comprehensive style guide for designing interruptive alerts (IAs) in EHR systems to improve clinician engagement and reduce alert fatigue that has been approved by our institutional alert governance committees. This style guide addresses critical aspects of IAs, including format, typography, color-coding, title brevity, patient identification, and introductory text. It also outlines the use of typographic emphasis, response options, default actions, and opt-out mechanisms, emphasizing the need for clear, concise, and actionable alerts that consider clinician workflow and cognitive burden.

Discussion: A standardized style guide for IAs can enhance clinician experience and clinical outcomes by reducing alert fatigue. Incorporating feedback and continuous evaluation of alert effectiveness is essential for maintaining relevance and supporting patient care within a dynamic clinical environment.

Background: The COVID pandemic accelerated the adoption of telehealth. Long-term implications for continued telehealth use on a large scale in primary care are still emerging, and the key to this understanding is how telehealth adoption impacts the delivery of health care.

Objectives: Our objective was to quantify how telehealth adoption during the pandemic impacted patients' access to healthcare and the usage patterns of different access modalities (in-office, audio, and video).

Methods: This study analyzed two years of de-identified EHR data from the ambulatory clinics at a Midwest medical school from 2020 and 2021. We focused on patient demographics, primary diagnoses, and patient preferences for receiving healthcare throughout the COVID-19 pandemic.

Results: Of the 105,362 patient visits analyzed across in-office, audio, and video visits, demographic data varied by gender, age, and racial composition. During the early pandemic, telehealth usage peaked in April 2020, with audio and video visits accounting for 45.8% and 18.07% of the total visits, respectively. These rates declined to averages of 12.47% and 3.58%, respectively, over the following months. Primary diagnoses during telehealth visits are often related to COVID-19 exposure or mental health or behavioral issues. Lastly, statistically significant associations exist between the visit modality and primary diagnoses.

Discussion: The COVID-19 pandemic initially saw a surge in audio telehealth visits; however, as safety measures were established, the mix of visit modalities returned to pre-lockdown levels. Video and audio telehealth maintained increased usage relative to pre-pandemic levels, potentially owing to increased awareness of the technology as a means for accessing healthcare and a learning curve associated with adoption. Patients preferred remote care during high COVID transmission. Video telehealth has shown significant associations with psychiatric, behavioral, and neurodevelopmental diagnoses. Disparities in video telehealth use among racial groups have highlighted potential access issues.

Background: Large language models (LLMs) have shown promise in various professional fields, including medicine and law. However, their performance in highly specialized tasks, such as extracting ICD-10-CM codes from patient notes, remains underexplored.

Objective: The primary objective was to evaluate and compare the performance of ICD-10-CM code extraction by different LLMs with that of human coder.

Methods: We evaluated performance of six LLMs (GPT-3.5, GPT-4, Claude 2.1, Claude 3, Gemini Advanced, and Llama 2-70b) in extracting ICD-10-CM codes against human coder. We used deidentified inpatient notes from American Health Information Management Association Vlab authentic patient cases for this study. We calculated percent agreement and Cohen's kappa values to assess the agreement between LLMs and human coder. We then identified reasons for discrepancies in code extraction by LLMs in a 10% random subset.

Results: Among 50 inpatient notes, human coder extracted 165 unique ICD-10-CM codes. LLMs extracted significantly higher number of unique ICD-10-CM codes than human coder, with Llama 2-70b extracting most (658) and Gemini Advanced the least (221). GPT-4 achieved highest percent agreement with human coder at 15.2%, followed by Claude 3 (12.7%) and GPT-3.5 (12.4%). Cohen's kappa values indicated minimal to no agreement, ranging from -0.02 to 0.01. When focusing on primary diagnosis, Claude 3 achieved highest percent agreement (26%) and kappa value (0.25). Reasons for discrepancies in extraction of codes varied amongst LLMs and included extraction of codes for diagnoses not confirmed by providers (60% with GPT-4), extraction of non-specific codes (25% with GPT-3.5), extraction of codes for signs and symptoms despite presence of more specific diagnosis (22% with Claude-2.1) and hallucinations (35% with Claude-2.1).

Conclusions: Current LLMs have poor performance in extraction of ICD-10-CM codes from inpatient notes when compared against the human coder.

Background: To keep pace with the developments in the medical informatics field, the curriculum of the Heidelberg/Heilbronn Medical Informatics Master of Science program is continuously updated. In its latest revision we restructured our program to allow more flexibility to accommodate updates and include current topics and to enable students' choices.

Objectives: To present our new concepts for graduate medical informatics education, share our experiences, and provide insights into the perception of these concepts by advanced students and graduates.

Methods: Our new curriculum consists of three core components: Areas of Concentration that bundle elective courses in an important domain of medical informatics, a large catalog of elective courses, introductory/alignment courses for students without a bachelor's degree in medical informatics. We conducted an online survey of graduates and students with at least 75 credits to assess their opinion on the program's effectiveness and attractiveness.

Results: Mandatory courses include clinical medicine, project management, research, and practical training in biomedical informatics. Five areas of concentration bundle elective courses for 30 credits to provide a solid foundation in an important domain in medical informatics. These are bioinformatics, data science, computer-aided diagnosis and therapy systems, information management, and software engineering in medicine. The catalog of electives offers a total of 67 courses. About 75% of them are assigned to more than one area of concentration. Our survey demonstrates that the participants highly appreciate the flexibility of the electives and the opportunity to develop an area of expertise.

Conclusion: Offering a high degree of flexibility to our students has motivated them to join our program and resulted in a high level of student satisfaction. By designing the curriculum with areas of concentration and providing an infrastructure that permits courses on emerging topics to be added easily to the curriculum, we were able to meet our students' expectations.

Background: Adolescents are at high risk for sexually transmitted infections (STIs) and frequently present to emergency departments (EDs) for care. Screening for STIs using confidential patient-reported outcomes represents an ideal use of electronic screening methodology.

Objectives: The objectives of this study were to implement a patient-facing, confidential electronic survey to assess adolescent risk for STIs and consent for testing with integrated provider facing electronic clinical decision support (CDS) across six geographically dispersed pediatric EDs and evaluate implementation based on survey and CDS usage metrics.

Methods: A pilot site provided code for the electronic survey, data query, and CDS templates to six EDs. Institutions identified necessary information technology (IT) personnel, completed local build, and made modifications to suit individual site workflow variations with all sites successfully deploying the electronic survey with electronic health record (EHR) -embedded CDS.

Results: 6,165 adolescents completed the confidential health survey between April 12, 2021 - September 25, 2022 out of 79,780 eligible adolescents. The CDS was triggered indicating the patient was at-risk or consented to STI testing across all six sites 2,058 times. The average percentage of time the CDS was acknowledged by a provider was 81.6% (range 45.7% - 97.6%). The median number of providers who acknowledged each instance of the CDS was 2.0. STI testing was ordered from the CDS on average 47.3% of the time. CDS acknowledge selections of "other" and "[testing] already ordered" were the most frequent indications STI testing was not ordered from the CDS.

Conclusions: Successful deployment of patient-facing screeners with integrated electronic CDS across multiple healthcare institutions is feasible. A combination of different types of IT and informatics expertise combined with local knowledge of clinical workflows is essential to success.

Background The Centers for Medicare & Medicaid Services (CMS) introduced changes in outpatient and inpatient evaluation and management (E/M) current procedural terminology (CPT) codes in 2021 and 2023, which were intended to streamline providers' clinical documentation. Objectives To study the effects of aligning inpatient and outpatient note templates with updated CMS guidelines on character length and documentation time per note at an internal medicine residency program in the southeastern United States. Methods In April 2023, the Atrium Health Wake Forest Baptist Internal Medicine Residency Program's inpatient and outpatient note templates were updated according to the most recent CMS guidelines. A pre-post analysis of resident documentation time and length was conducted comparing notes written with the residency note templates from May 1, 2022 to August 31, 2022 (6439 notes) to notes written with the residency note templates from May 1, 2023 to August 31, 2023 (8828 notes). Interns were surveyed regarding their perceptions of the updated note templates. Results After the note template updates, when adjusted for differing percentages of note types in the pre- and post-periods and accounting for multiple notes written by each resident, notes written with the residency note templates decreased by a mean character length of -882 characters (95% CI: -953, -811, p<.0001), while time spent writing notes did not significantly decrease. 17/17 respondents had favorable perceptions of the note templates. Conclusions The internal medicine residency inpatient and outpatient note templates were updated to align with the most recent CMS E/M documentation requirement changes. These note template changes were associated with a meaningful decrease in documentation length but no overall significant reduction in mean documentation time when adjusted for differing percentages of note types in the pre- and post-periods and multiple notes written by the same author. The interns perceived the note template changes positively.

Objective: To assist residents in selecting the correct Current Procedural Terminology (CPT) code for evaluation and management (E/M) services thru the addition of disappearing help text into a standardized note template.

Methods: We created a disappearing text block that summarizes E/M requirements and embedded it into the note template used by residents at a pediatric urgent care clinic. An intervention cohort composed of post graduate year 1 (PGY 1) residents was instructed to use this note template, while senior residents (PGY 2-3) were instructed to use an identical template that lacked the help text. We evaluated the incidence of CPT change by the attending physician for each visit as a proxy for improvement in resident billing practices. Logistic regression with a primary outcome of whether the encounter CPT code was changed was completed.

Results: There were 2,869 encounters during the 255-day study period; the help text was used in 1,112 (38.8%) encounters. There was some crossover in note use; i.e., PGY 1s using the note without help text and PGY 2s using the note with help text. Nevertheless, all residents who used the help text had a lower unadjusted rate of CPT change (22.1% vs 30.6%, OR= 0.64, p < 0.0001). This pattern persisted when stratified by trainee level - PGY 1 (22.6% vs 45.3%, OR=0.35,p < 0.0001) and PGY 2-3 (12.2% vs 27.8%, p = 0.018). Adjusting for multiple factors, the use of help text was associated with a lower incidence of CPT change (odds ratio [OR] = 0.28, 95% confidence interval [CI]: 0.19-0.44).

Conclusions: Residents' use of the disappearing help text was associated with a large decrease in CPT code adjustment by attending physicians, which demonstrates its promise for improved E/M coding and for other applications.

The NIH Pragmatic Trials Collaboratory supports the design and conduct of 31 embedded pragmatic clinical trials, and many of these trials use patient-reported outcome measures (PROMs) to provide valuable information about their patients' health and wellness. Often these trials enroll medically underserved patients, including people with incomes below the federal poverty threshold, racial or ethnically minoritized groups, or rural or frontier communities. In this series of trial case reports, we provide lessons learned about collecting PROMs in these populations. The unbiased collection of PROM data is critical to increase the generalizability of trial outcomes and to address health inequities. Use of electronic health records (EHRs) and other digital modes of PROM administration have gained traction. However, engagement with these modes is often low among disparities prone populations due to lessened digital proficiency, device access, and uptake of EHR portals and web interfaces. To maximize the completeness and representativeness of their trial outcome data, study teams tested a range of strategies to improve PROM response rates with emphasis on disparities prone and underserved patient groups. This manuscript describes the approaches, their implementation, and the targeted populations. Optimized PROM collection required hybrid approaches with multiple outreach modes, high-touch methods, creativity in promoting digital uptake, multi-modal participant engagement, and text messaging.

Background: Healthcare provider burnout is a critical issue with significant implications for individual well-being, patient care, and healthcare system efficiency. Addressing burnout is essential for improving both provider well-being and the quality of patient care. Ambient artificial intelligence (AI) offers a novel approach to mitigating burnout by reducing the documentation burden through advanced speech recognition and natural language processing technologies that summarize the patient encounter into a clinical note to be reviewed by clinicians.

Objective: To assess provider burnout and professional fulfilment associated with Ambient AI technology during a pilot study, assessed using the Stanford Professional Fulfillment Index (PFI).

Methods: A pre-post observational study was conducted at University of Iowa Health Care with 38 volunteer physicians and advanced practice providers. Participants used a commercial ambient AI tool, over a 5-week trial in ambulatory environments. The AI tool transcribed patient-clinician conversations and generated preliminary clinical notes for review and entry into the electronic medical record. Burnout and professional fulfillment were assessed using the Stanford PFI at baseline and post-intervention.

Results: Pre-test and post-test surveys were completed by 35/38 participants (92% survey completion rate). Results showed a significant reduction in burnout scores, with the median burnout score improving from 4.16 to 3.16 (p=0.005), with validated Stanford PFI cutoff for overall burnout 3.33. Burnout rates decreased from 69% to 43%. There was a notable improvement in interpersonal disengagement scores (3.6 vs. 2.5, p<0.001), although work exhaustion scores did not significantly change. Professional fulfillment showed a modest, non-significant upward trend (6.1 vs. 6.5, p=0.10).

Conclusions: Ambient AI significantly reduces healthcare provider burnout and may enhance professional fulfillment. By alleviating documentation burdens, ambient AI can improve operational efficiency and provider well-being. These findings suggest that broader implementation of ambient AI could be a strategic intervention to combat burnout in healthcare settings.