Examining Linguistic Differences in Electronic Health Records for Diverse Patients With Diabetes: Natural Language Processing Analysis.

Background: Individuals from minoritized racial and ethnic backgrounds experience pernicious and pervasive health disparities that have emerged, in part, from clinician bias. Objective: We used a natural language processing approach to examine whether linguistic markers in electronic health record (EHR) notes differ based on the race and ethnicity of the patient. To validate this methodological approach, we also assessed the extent to which clinicians perceive linguistic markers to be indicative of bias. Methods: In this cross-sectional study, we extracted EHR notes for patients who were aged 18 years or older; had more than 5 years of diabetes diagnosis codes; and received care between 2006 and 2014 from family physicians, general internists, or endocrinologists practicing in an urban, academic network of clinics. The race and ethnicity of patients were defined as White non-Hispanic, Black non-Hispanic, or Hispanic or Latino. We hypothesized that Sentiment Analysis and Social Cognition Engine (SEANCE) components (ie, negative adjectives, positive adjectives, joy words, fear and disgust words, politics words, respect words, trust verbs, and well-being words) and mean word count would be indicators of bias if racial differences emerged. We performed linear mixed effects analyses to examine the relationship between the outcomes of interest (the SEANCE components and word count) and patient race and ethnicity, controlling for patient age. To validate this approach, we asked clinicians to indicate the extent to which they thought variation in the use of SEANCE language domains for different racial and ethnic groups was reflective of bias in EHR notes. Results: We examined EHR notes (n=12,905) of Black non-Hispanic, White non-Hispanic, and Hispanic or Latino patients (n=1562), who were seen by 281 physicians. A total of 27 clinicians participated in the validation study. In terms of bias, participants rated negative adjectives as 8.63 (SD 2.06), fear and disgust words as 8.11 (SD 2.15), and positive adjectives as 7.93 (SD 2.46) on a scale of 1 to 10, with 10 being extremely indicative of bias. Notes for Black non-Hispanic patients contained significantly more negative adjectives (coefficient 0.07, SE 0.02) and significantly more fear and disgust words (coefficient 0.007, SE 0.002) than those for White non-Hispanic patients. The notes for Hispanic or Latino patients included significantly fewer positive adjectives (coefficient -0.02, SE 0.007), trust verbs (coefficient -0.009, SE 0.004), and joy words (coefficient -0.03, SE 0.01) than those for White non-Hispanic patients. Conclusions: This approach may enable physicians and researchers to identify and mitigate bias in medical interactions, with the goal of reducing health disparities stemming from bias.

Racial/Ethnic Minority Identifying Family Physicians Are More Likely to Work in Solo Practices.

While the overall proportion of family physicians who work in solo practices has been steadily declining, Black, Hispanic/Latino, and Asian family physicians are more likely to work in these settings. Given their association with high levels of continuity and improved health outcomes, and given patient preference for racial concordance with their physicians, policy makers and payors should consider how to support family physicians in solo practice in the interest of promoting access to and quality of care for ethnic/racial minorities.

An Assessment of How Clinicians and Staff Members Use a Diabetes Artificial Intelligence Prediction Tool: Mixed Methods Study.

BACKGROUND: Nearly one-third of patients with diabetes are poorly controlled (hemoglobin A1c≥9%). Identifying at-risk individuals and providing them with effective treatment is an important strategy for preventing poor control. OBJECTIVE: This study aims to assess how clinicians and staff members would use a clinical decision support tool based on artificial intelligence (AI) and identify factors that affect adoption. METHODS: This was a mixed methods study that combined semistructured interviews and surveys to assess the perceived usefulness and ease of use, intent to use, and factors affecting tool adoption. We recruited clinicians and staff members from practices that manage diabetes. During the interviews, participants reviewed a sample electronic health record alert and were informed that the tool uses AI to identify those at high risk for poor control. Participants discussed how they would use the tool, whether it would contribute to care, and the factors affecting its implementation. In a survey, participants reported their demographics; rank-ordered factors influencing the adoption of the tool; and reported their perception of the tool's usefulness as well as their intent to use, ease of use, and organizational support for use. Qualitative data were analyzed using a thematic content analysis approach. We used descriptive statistics to report demographics and analyze the findings of the survey. RESULTS: In total, 22 individuals participated in the study. Two-thirds (14/22, 63%) of respondents were physicians. Overall, 36% (8/22) of respondents worked in academic health centers, whereas 27% (6/22) of respondents worked in federally qualified health centers. The interviews identified several themes: this tool has the potential to be useful because it provides information that is not currently available and can make care more efficient and effective; clinicians and staff members were concerned about how the tool affects patient-oriented outcomes and clinical workflows; adoption of the tool is dependent on its validation, transparency, actionability, and design and could be increased with changes to the interface and usability; and implementation would require buy-in and need to be tailored to the demands and resources of clinics and communities. Survey findings supported these themes, as 77% (17/22) of participants somewhat, moderately, or strongly agreed that they would use the tool, whereas these figures were 82% (18/22) for usefulness, 82% (18/22) for ease of use, and 68% (15/22) for clinic support. The 2 highest ranked factors affecting adoption were whether the tool improves health and the accuracy of the tool. CONCLUSIONS: Most participants found the tool to be easy to use and useful, although they had concerns about alert fatigue, bias, and transparency. These data will be used to enhance the design of an AI tool.

Primary Care: The Actual Intelligence Required for Artificial Intelligence to Advance Health Care and Improve Health.

With conversational agents triaging symptoms, cameras aiding diagnoses, and remote sensors monitoring vital signs, the use of artificial intelligence (AI) outside of hospitals has the potential to improve health, according to a recently released report from the National Academy of Medicine. Despite this promise, the success of AI is not guaranteed, and stakeholders need to be involved with its development to ensure that the resulting tools can be easily used by clinicians, protect patient privacy, and enhance the value of the care delivered. A crucial stakeholder group missing from the conversation is primary care. As the nation's largest delivery platform, primary care will have a powerful impact on whether AI is adopted and subsequently exacerbates health disparities. To leverage these benefits, primary care needs to serve as a medical home for AI, broaden its teams and training, and build on government initiatives and funding.

How do aggression source, employee characteristics and organisational response impact the relationship between workplace aggression and work and health outcomes in healthcare employees? A cross-sectional analysis of the National Health Service staff survey in England.

OBJECTIVES: To examine the prevalence of aggression in healthcare and its association with employees' turnover intentions, health and engagement, as well as how these effects differ based on aggression source (patients vs colleagues), employee characteristics (race, gender and occupation) and organisational response to the aggression. DESIGN: Multilevel moderated regression analysis of 2010 National Health Service (NHS) survey. SETTING: 147 acute NHS trusts in England. PARTICIPANTS: 36 850 participants across three occupational groups (14% medical/dental, 61% nursing/midwifery, 25% allied health professionals or scientific and technical staff). MAIN OUTCOME MEASURES: Employee turnover intentions, health and work engagement. RESULTS: Both forms of aggression (from patients and colleagues) have significant and substantial effects on turnover intentions, health and work engagement; however, for all three outcome variables, the effect of aggression from colleagues is more than twice the size of the effect of aggression from patients. Organisational response was found to buffer the negative effects of aggression from patients for turnover intentions and the negative effects of aggression from patients and colleagues for employee health. The results also demonstrated that nurses/midwives, women and Black employees are more likely to experience aggression; however, no clear patterns emerged on how aggression differentially impacts employees of different races, genders and occupations with respect to the outcome variables. CONCLUSIONS: Although aggression from patients and colleagues both have negative effects on healthcare employees' turnover intentions, health and work engagement, these negative effects are worse when it is aggression from colleagues. Having an effective organisational response can help ameliorate the negative effects of aggression on employees' health; however, it may not always buffer negative effects on turnover intentions and work engagement. Future research should examine other approaches, as well as how organisational responses and resources may need to differ based on aggression source.

Enhanced care planning and clinical-community linkages versus usual care to address basic needs of patients with multiple chronic conditions: a clinician-level randomized controlled trial.

BACKGROUND: Many patients with poorly controlled multiple chronic conditions (MCC) also have unhealthy behaviors, mental health challenges, and unmet social needs. Medical management of MCC may have limited benefit if patients are struggling to address their basic life needs. Health systems and communities increasingly recognize the need to address these issues and are experimenting with and investing in new models for connecting patients with needed services. Yet primary care clinicians, whose regular contact with patients makes them more familiar with patients' needs, are often not included in these systems. METHODS: We are starting a clinician-level cluster-randomized controlled trial to evaluate how primary care clinicians can participate in these community and hospital solutions and whether doing so is effective in controlling MCC. Sixty clinicians in the Virginia Ambulatory Care Outcomes Research Network will be matched by age and sex and randomized to usual care (control condition) or enhanced care planning with clinical-community linkage support (intervention). From the electronic health record we will identify all patients with MCC, including cardiovascular disease or risks, diabetes, obesity, or depression. A baseline assessment will be mailed to up to 50 randomly selected patients for each clinician (3000 total). Ten respondents per clinician (600 patients total) with uncontrolled MCC will be randomly selected for study inclusion, with oversampling of minorities. The intervention includes two components. First, we will use an enhanced care planning tool, My Own Health Report (MOHR), to screen patients for health behavior, mental health, and social needs. Patients will be supported by a patient navigator, who will help patients prioritize needs, create care plans, and write a personal narrative to guide the care team. Patients will update care plans every 1 to 2 weeks. Second, we will create community-clinical linkage to help address patients' needs. The linkage will include community resource registries, personnel to span settings (patient navigators and a community health worker), and care team coordination across team members through MOHR. DISCUSSION: This study will help inform efforts by primary care clinicians to help address unhealthy behaviors, mental health needs, and social risks as a strategy to better control MCC. TRIAL REGISTRATION: ClinicalTrials.gov: NCT03885401. Registered on 19 September 2019.

Clinicians' Overestimation of Their Geographic Service Area.

In this study, we evaluated family physicians' ability to estimate the service area of their patient panel-a critical first step in contextual population-based primary care. We surveyed 14 clinicians and administrators from 6 practices. Participants circled their estimated service area on county maps that were compared with the actual service area containing 70% of the practice's patients. Accuracy was ascertained from overlap and the amount of estimated census tracts that were not part of the actual service area. Average overlap was 75%, but participants overestimated their service area by an average of 166 square miles. Service area overestimation impedes implementation of targeted community interventions by practices.

Loneliness in Primary Care Patients: A Prevalence Study.

PURPOSE: Loneliness has important health consequences. Little is known, however, about loneliness in primary care patient populations. This study describes the prevalence of loneliness in patients presenting for primary care and associations with self-reported demographic factors, health care utilization, and health-related quality of life. METHODS: We conducted cross-sectional surveys of adults presenting for routine care to outpatient primary care practices in 2 diverse practice-based research networks. The 3-item University of California, Los Angeles Loneliness Scale was utilized to determine loneliness. RESULTS: The prevalence of loneliness was 20% (246/1,235). Loneliness prevalence was inversely associated with age (P <.01) and less likely in those who were married (P <.01) or employed (P <.01). Loneliness was more common in those with lower health status (P <.01), including when adjusting for employment and relationship status (odds ratio [OR] = 1.05; 95% CI, 1.03-1.07). Primary care visits (OR = 1.07; 95% CI, 1.03-1.10), urgent care/emergency department visits (OR = 1.24; 95% CI, 1.12-1.38), and hospitalizations (OR = 1.15; 95% CI, 1.01-1.31) were associated with loneliness status. There was no significant difference in rates of loneliness between sexes (P = .08), racial categories (P = .57), or rural and urban respondents (P = .42). CONCLUSIONS: Our findings demonstrate that loneliness is common in primary care patients and is associated with adverse health consequences including poorer health status and greater health care utilization. Further work is needed to understand the value of screening for and using interventions to treat loneliness in primary care.

Geographic Characteristics of Loneliness in Primary Care.

Loneliness is associated with poor health outcomes, and there is growing attention on loneliness as a social determinant of health. Our study sought to determine the associations between community factors and loneliness. The Three-Item Loneliness Scale and zip codes of residence were collected in primary care practices in Colorado and Virginia. Living in zip codes with higher unemployment, poor access to health care, lower income, higher proportions of blacks, and poor transportation was associated with higher mean loneliness scores. Future studies that examine interventions addressing loneliness may be more effective if they consider social context and community characteristics.

Disconnected: a survey of users and nonusers of telehealth and their use of primary care.

OBJECTIVE: The study sought to assess awareness, perceptions, and value of telehealth in primary care from the perspective of patients. MATERIALS AND METHODS: We conducted a cross-sectional, Web-based survey of adults with access to telehealth services who visited healthcare providers for any of the 20 most-commonly seen diagnoses during telehealth visits. Three groups were studied: registered users (RUs) of telehealth had completed a LiveHealth Online (a health plan telehealth service provider) visit, registered nonusers (RNUs) registered for LiveHealth Online but had not conducted a visit, and nonregistered nonusers (NRNUs) completed neither step. RESULTS: Of 32 831 patients invited, 3219 (9.8%) responded and 766 met eligibility criteria and completed surveys: 390 (51%) RUs, 117 (15%) RNUs, and 259 (34%) NRNUs. RUs were least likely to have a primary care usual source of care (65.6% vs 78.6% for RNUs vs 80.0% for NRNUs; P < .001). Nearly half (46.8%) of RUs were unable to get an appointment with their doctor, and 34.8% indicated that their doctor's office was closed. Among the 3 groups, RUs were most likely to be employed (89.5% vs 88.9% vs 82.2%; P = .007), have post-high school education (94.4% vs 93.2% vs 86.5%; P = .003), and live in urban areas (81.0% vs 69.2% vs 76.0%; P = .021). CONCLUSIONS: Telehealth users reported that they relied on live video for enhanced access and were less connected to primary care than nonusers were. Telehealth may expand service access but risks further fragmentation of care and undermining of the primary care function absent better coordination and information sharing with usual sources of patients' care.

Clinician Experiences with Screening for Social Needs in Primary Care.

BACKGROUND: Despite clear evidence demonstrating the influence of social determinants on health, whether and how clinicians should address these determinants remain unclear. We aimed to understand primary care clinicians' experiences of administering a social needs screening instrument. METHODS: Using a prospective, observational design, we identified patients living in communities with lower education and income seen by 17 clinicians from 12 practices in northern Virginia. Before office visits, patients completed social needs surveys, which probed about their quality of life, education, housing, finances, substance use, transportation, social connections, physical activity, and food access. Clinicians then reviewed the completed surveys with patients. Concurrently, clinicians participated in a series of learning collaboratives to consider how to address social needs as part of care and completed diary entries about how knowing the patient's social needs influenced care after seeing each patient. RESULTS: Out of a total of 123 patients, 106 (86%) reported a social need. Excluding physical activity, 71% reported a social need, although only 3% wanted help. Clinicians reported that knowing the patient had a social need changed care delivery in 23% of patients and helped improve interactions with and knowledge of the patient in 53%. Clinicians reported that assessing social needs is difficult and resource intensive and that there were insufficient resources to help patients with identified needs. CONCLUSIONS: Clinicians reported that knowing patients' social needs changed what they did and improved communication for many patients. However, more evidence is needed regarding the benefit of social needs screening in primary care before widespread implementation.

Increasing Family Medicine Faculty Diversity Still Lags Population Trends.

BACKGROUND: Faculty diversity has important implications for medical student diversity. The purpose of this analysis is to describe trends in racial, ethnic, and gender diversity in family medicine (FM) departments and compare these trends to the diversity of matriculating medical students, the diversity of all medical school faculty, and the population in general. METHODS: We used the Association of American Medical Colleges Faculty Roster to describe trends in proportions of female and minorities under-represented in medicine (URM) in FM department full-time faculty in U.S. MD-granting medical schools. RESULTS: Among FM faculty, the proportions of female and URM faculty have grown more than 2-fold between 1980 and 2015. Increasing faculty rank was associated with lower diversity across the study period. FM departments had higher female and URM proportions than the average of all other specialties, but URM representation still lagged population trends. CONCLUSION: Although FM faculty diversity is growing over time, continued attention to URM representation should remain a priority.

When Do Primary Care Physicians Retire? Implications for Workforce Projections.

PURPOSE: Retirement of primary care physicians is a matter of increasing concern in light of physician shortages. The joint purposes of this investigation were to identify the ages when the majority of primary care physicians retire and to compare this with the retirement ages of practitioners in other specialties. METHODS: This descriptive study was based on AMA Physician Masterfile data from the most recent 5 years (2010-2014). We also compared 2008 Masterfile data with data from the National Plan and Provider Enumeration System to calculate an adjustment for upward bias in retirement ages when using the Masterfile alone. The main analysis defined retirement as leaving clinical practice. The primary outcome was construction of a retirement curve. Secondary outcomes involved comparisons of retirement interquartile ranges (IQRs) by sex and practice location across specialties. RESULTS: The 2014 Masterfile included 77,987 clinically active primary care physicians between ages 55 and 80 years. The median age of retirement from clinical activity of all primary care physicians who retired in the period from 2010 to 2014 was 64.9 years, (IQR, 61.4-68.3); the median age of retirement from any activity was 66.1 years (IQR, 62.6-69.5). However measured, retirement ages were generally similar across primary care specialties. Females had a median retirement about 1 year earlier than males. There were no substantive differences in retirement ages between rural and urban primary care physicians. CONCLUSIONS: Primary care physicians in our data tended to retire in their mid-60s. Relatively small differences across sex, practice location, and time suggest that changes in the composition of the primary care workforce will not have a remarkable impact on overall retirement rates in the near future.

Solo and Small Practices: A Vital, Diverse Part of Primary Care.

PURPOSE: Solo and small practices are facing growing pressure to consolidate. Our objectives were to determine (1) the percentage of family physicians in solo and small practices, and (2) the characteristics of and services provided by these practices. METHODS: A total of 10,888 family physicians seeking certification through the American Board of Family Medicine in 2013 completed a demographic survey. Their practices were split into categories by size: solo, small (2 to 5 providers), medium (6 to 20 providers), and large (more than 20 providers). We also determined the rurality of the county where the physicians practiced. We developed 2 logistic regression models: one assessed predictors of practicing in a solo or small practice, while the other was restricted to solo and small practices and assessed predictors of practicing in a solo practice. RESULTS: More than one-half of respondents worked in solo or small practices. Small practices were the largest group (36%) and were the most likely to be located in a rural setting (20%). The likelihood of having a care coordinator and medical home certification increased with practice size. Physicians were more likely to be practicing in small or solo practices (vs medium-sized or large ones) if they were African American or Hispanic, had been working for more than 30 years, and worked in rural areas. Physicians were more likely to be practicing in small practices (vs solo ones) if they worked in highly rural areas. CONCLUSIONS: Family physicians in solo and small practices comprised the majority among all family physicians seeking board certification and were more likely to work in rural geographies. Extension programs and community health teams have the potential to support transformation within these practices.

Estimating the residency expansion required to avoid projected primary care physician shortages by 2035.

PURPOSE: The purpose of this study was to calculate the projected primary care physician shortage, determine the amount and composition of residency growth needed, and estimate the impact of retirement age and panel size changes. METHODS: We used the 2010 National Ambulatory Medical Care Survey to calculate utilization of ambulatory primary care services and the US Census Bureau to project demographic changes. To determine the baseline number of primary care physicians and the number retiring at 66 years, we used the 2014 American Medical Association Masterfile. Using specialty board and American Osteopathic Association figures, we estimated the annual production of primary care residents. To calculate shortages, we subtracted the accumulated primary care physician production from the accumulated number of primary care physicians needed for each year from 2015 to 2035. RESULTS: More than 44,000 primary care physicians will be needed by 2035. Current primary care production rates will be unable to meet demand, resulting in a shortage in excess of 33,000 primary care physicians. Given current production, an additional 1,700 primary care residency slots will be necessary by 2035. A 10% reduction in the ratio of population per primary care physician would require more than 3,000 additional slots by 2035, whereas changing the expected retirement age from 66 years to 64 years would require more than 2,400 additional slots. CONCLUSIONS: To eliminate projected shortages in 2035, primary care residency production must increase by 21% compared with current production. Delivery models that shift toward smaller ratios of population to primary care physicians may substantially increase the shortage.

Cost, utilization, and quality of care: an evaluation of illinois' medicaid primary care case management program.

PURPOSE: In 2006, Illinois established Illinois Health Connect (IHC), a primary care case management program for Medicaid that offered enhanced fee-for-service, capitation payments, performance incentives, and practice support. Illinois also implemented a complementary disease management program, Your Healthcare Plus (YHP). This external evaluation explored outcomes associated with these programs. METHODS: We analyzed Medicaid claims and enrollment data from 2004 to 2010, covering both pre- and post-implementation. The base year was 2006, and 2006-2010 eligibility criteria were applied to 2004-2005 data to allow comparison. We studied costs and utilization trends, overall and by service and setting. We studied quality by incorporating Healthcare Effectiveness Data and Information Set (HEDIS) measures and IHC performance payment criteria. RESULTS: Illinois Medicaid expanded considerably between 2006 (2,095,699 full-year equivalents) and 2010 (2,692,123). Annual savings were 6.5% for IHC and 8.6% for YHP by the fourth year, with cumulative Medicaid savings of $1.46 billion. Per-beneficiary annual costs fell in Illinois over this period compared to those in states with similar Medicaid programs. Quality improved for nearly all metrics under IHC, and most prevention measures more than doubled in frequency. Medicaid inpatient costs fell by 30.3%, and outpatient costs rose by 24.9% to 45.7% across programs. Avoidable hospitalizations fell by 16.8% for YHP, and bed-days fell by 15.6% for IHC. Emergency department visits declined by 5% by 2010. CONCLUSIONS: The Illinois Medicaid IHC and YHP programs were associated with substantial savings, reductions in inpatient and emergency care, and improvements in quality measures. This experience is not typical of other states implementing some, but not all, of these same policies. Although specific features of the Illinois reforms may have accounted for its better outcomes, the limited evaluation design calls for caution in making causal inferences.

Clinical and community delivery systems for preventive care: an integration framework.

Although clinical preventive services (CPS)-screening tests, immunizations, health behavior counseling, and preventive medications-can save lives, Americans receive only half of recommended services. This "prevention gap," if closed, could substantially reduce morbidity and mortality. Opportunities to improve delivery of CPS exist in both clinical and community settings, but these activities are rarely coordinated across these settings, resulting in inefficiencies and attenuated benefits. Through a literature review, semi-structured interviews with 50 national experts, field observations of 53 successful programs, and a national stakeholder meeting, a framework to fully integrate CPS delivery across clinical and community care delivery systems was developed. The framework identifies the necessary participants, their role in care delivery, and the infrastructure, support, and policies necessary to ensure success. Essential stakeholders in integration include clinicians; community members and organizations; spanning personnel and infrastructure; national, state, and local leadership; and funders and purchasers. Spanning personnel and infrastructure are essential to bring clinicians and communities together and to help patients navigate across care settings. The specifics of clinical-community integrations vary depending on the services addressed and the local context. Although broad establishment of effective clinical-community integrations will require substantial changes, existing clinical and community models provide an important starting point. The key policies and elements of the framework are often already in place or easily identified. The larger challenge is for stakeholders to recognize how integration serves their mutual interests and how it can be financed and sustained over time.

Projecting US primary care physician workforce needs: 2010-2025.

PURPOSE: We sought to project the number of primary care physicians required to meet US health care utilization needs through 2025 after passage of the Affordable Care Act. METHODS: In this projection of workforce needs, we used the Medical Expenditure Panel Survey to calculate the use of office-based primary care in 2008. We used US Census Bureau projections to account for demographic changes and the American Medical Association's Masterfile to calculate the number of primary care physicians and determine the number of visits per physician. The main outcomes were the projected number of primary care visits through 2025 and the number of primary care physicians needed to conduct those visits. RESULTS: Driven by population growth and aging, the total number of office visits to primary care physicians is projected to increase from 462 million in 2008 to 565 million in 2025. After incorporating insurance expansion, the United States will require nearly 52,000 additional primary care physicians by 2025. Population growth will be the largest driver, accounting for 33,000 additional physicians, while 10,000 additional physicians will be needed to accommodate population aging. Insurance expansion will require more than 8,000 additional physicians, a 3% increase in the current workforce. CONCLUSIONS: Population growth will be the greatest driver of expected increases in primary care utilization. Aging and insurance expansion will also contribute to utilization, but to a smaller extent.