Using Predictive Analytics to Guide Patient Care and Research in a National Health System

Rates of diagnosis and service utilization in veterans with hoarding disorder

Hoarding Disorder (HD) is a prominent and disabling neuropsychiatric condition defined by the inability to discard objects resulting in impairing levels of clutter. The prevalence rate is 2-6 % and increases with age. The aging Veteran population is a high risk group for impairment associated with HD. Medical and psychiatric comorbidities as well as associated rates of disability and poor quality of life are very common in both HD and the related disorder of OCD. We examined rates of HD and OCD diagnoses at the VA San Diego Healthcare System. Data were obtained from medical records for all Veterans with these diagnoses over 8-years and included information on medical and psychiatric care, homelessness services, and Care Assessment Needs (CAN) scores. Rates of diagnosis for both HD and OCD were well below epidemiological estimates. Veterans with HD were older, had higher rates of medical hospital admissions with longer stays; had more cardiac, neurological, and acquired medical conditions; had more psychiatric comorbidities; had more interactions with the suicide prevent team and homelessness services; and had higher CAN scores than Veterans with OCD. The low rate of diagnosis and high services utilization of Veterans with HD demonstrates an area of unmet need.

Variation in benefit among patients with serious mental illness who receive integrated psychiatric and primary care

PURPOSE

The population with serious mental illness has high risk for hospitalization or death due to unhealthy behaviors and inadequate medical care, though the level of risk varies substantially. Programs that integrate medical and psychiatric services improve outcomes but are challenging to implement and access is limited. It would be useful to know whether benefits are confined to patients with specific levels of risk.

METHODS

In a population with serious mental illness and increased risk for hospitalization or death, a specialized medical home integrated services and improved treatment and outcomes. Treatment quality, chronic illness care, care experience, symptoms, and quality of life were assessed for a median of 385 days. Analyses examine whether improvements varied by baseline level of patient risk.

RESULTS

Patients with greater risk were more likely to be older, more cognitively impaired, and have worse mental health. Integrated services increased appropriate screening for body mass index, lipids, and glucose, but increases did not differ significantly by level of risk. Integrated services also improved chronic illness care, care experience, mental health-related quality of life, and psychotic symptoms. There were also no significant differences by risk level.

CONCLUSIONS

There were benefits from integration of primary care and psychiatric care at all levels of increased risk, including those with extremely high risk above the 95th percentile. When developing integrated care programs, patients should be considered at all levels of risk, not only those who are the healthiest.

The Characteristics of People with Serious Mental Illness Who are at High Risk for Hospitalization or Death

Many individuals with serious mental illness are at high risk for hospitalization or death due to inadequate treatment of medical conditions or unhealthy behaviors. The authors describe demographic and clinical characteristics associated with increased risk in this population. Electronic data were obtained for individuals in treatment at a large Veterans' healthcare system who were at high risk according to a validated model. A random sample of these individuals was assessed in person. Multivariable regressions estimated the effect of numerous demographic, health, and clinical characteristics on risk. Emergency visits and hospitalizations were common. Greater risk was associated with being male, not married, and having more diagnoses. While risk varied by race, this effect was no longer significant after controlling for other factors. Health-related quality of life worsened with increasing risk. Routine data identify a large population of high-risk individuals who may benefit from outreach to provide healthcare services.

Strategies to Improve Perioperative Palliative Care Integration for Seriously Ill Veterans

CONTEXT

Seriously ill patients are at higher risk for adverse surgical outcomes. Palliative care (PC) interventions for seriously ill surgical patients are associated with improved quality of patient care and patient-centered outcomes, yet, they are underutilized perioperatively.

OBJECTIVES

To identify strategies for improving perioperative PC integration for seriously ill Veterans from the perspectives of PC providers and surgeons.

METHODS

We conducted semistructured, in-depth individual and group interviews with Veteran Health Administration PC team members and surgeons between July 2020 and April 2021. Participants were purposively sampled from high- and low-collaboration sites based on the proportion of received perioperative palliative consults. We performed a team-based thematic analysis with dual coding (inter-rater reliability above 0.8).

RESULTS

Interviews with 20 interdisciplinary PC providers and 13 surgeons at geographically distributed Veteran Affairs sites converged on four strategies for improving palliative care integration and goals of care conversations in the perioperative period: 1) develop and maintain collaborative, trusting relationships between palliative care providers and surgeons; 2) establish risk assessment processes to identify patients who may benefit from a PC consult; 3) involve both PC providers and surgeons at the appropriate time in the perioperative workflow; 4) provide sufficient resources to allow for an interdisciplinary sharing of care.

CONCLUSION

The study demonstrates that individual, programmatic, and organizational efforts could facilitate interservice collaboration between PC clinicians and surgeons.

Eliminating Algorithmic Racial Bias in Clinical Decision Support Algorithms: Use Cases from the Veterans Health Administration

The Veterans Health Administration uses equity- and evidence-based principles to examine, correct, and eliminate use of potentially biased clinical equations and predictive models. We discuss the processes, successes, challenges, and next steps in four examples. We detail elimination of the race modifier for estimated kidney function and discuss steps to achieve more equitable pulmonary function testing measurement. We detail the use of equity lenses in two predictive clinical modeling tools: Stratification Tool for Opioid Risk Mitigation (STORM) and Care Assessment Need (CAN) predictive models. We conclude with consideration of ways to advance racial health equity in clinical decision support algorithms.

Revolutionizing healthcare: the role of artificial intelligence in clinical practice

INTRODUCTION

Healthcare systems are complex and challenging for all stakeholders, but artificial intelligence (AI) has transformed various fields, including healthcare, with the potential to improve patient care and quality of life. Rapid AI advancements can revolutionize healthcare by integrating it into clinical practice. Reporting AI's role in clinical practice is crucial for successful implementation by equipping healthcare providers with essential knowledge and tools.

RESEARCH SIGNIFICANCE

This review article provides a comprehensive and up-to-date overview of the current state of AI in clinical practice, including its potential applications in disease diagnosis, treatment recommendations, and patient engagement. It also discusses the associated challenges, covering ethical and legal considerations and the need for human expertise. By doing so, it enhances understanding of AI's significance in healthcare and supports healthcare organizations in effectively adopting AI technologies.

MATERIALS AND METHODS

The current investigation analyzed the use of AI in the healthcare system with a comprehensive review of relevant indexed literature, such as PubMed/Medline, Scopus, and EMBASE, with no time constraints but limited to articles published in English. The focused question explores the impact of applying AI in healthcare settings and the potential outcomes of this application.

RESULTS

Integrating AI into healthcare holds excellent potential for improving disease diagnosis, treatment selection, and clinical laboratory testing. AI tools can leverage large datasets and identify patterns to surpass human performance in several healthcare aspects. AI offers increased accuracy, reduced costs, and time savings while minimizing human errors. It can revolutionize personalized medicine, optimize medication dosages, enhance population health management, establish guidelines, provide virtual health assistants, support mental health care, improve patient education, and influence patient-physician trust.

CONCLUSION

AI can be used to diagnose diseases, develop personalized treatment plans, and assist clinicians with decision-making. Rather than simply automating tasks, AI is about developing technologies that can enhance patient care across healthcare settings. However, challenges related to data privacy, bias, and the need for human expertise must be addressed for the responsible and effective implementation of AI in healthcare.

Association Between Mental Health Conditions and Outpatient Care Fragmentation: a National Study of Older High-Risk Veterans

BACKGROUND

Healthcare fragmentation may lead to adverse consequences and may be amplified among older, sicker patients with mental health (MH) conditions.

OBJECTIVE

To determine whether older Veterans with MH conditions have more fragmented outpatient non-MH care, compared with older Veterans with no MH conditions.

DESIGN

Retrospective cohort study using FY2014 Veterans Health Administration (VHA) administrative data linked to Medicare data.

PARTICIPANTS

125,481 VHA patients ≥ 65 years old who were continuously enrolled in Medicare Fee-for-Service Parts A and B and were at high risk for hospitalization.

MAIN OUTCOME AND MEASURES

The main outcome was non-MH care fragmentation as measured by (1) non-MH provider count and (2) Usual Provider of Care (UPC), the proportion of care with the most frequently seen non-MH provider. We tested the association between no vs. any MH conditions and outcomes using Poisson regression and fractional regression with logit link, respectively. We also compared Veterans with no MH condition with each MH condition and combinations of MH conditions, adjusting for sociodemographics, comorbidities, and drive-time to VHA specialty care.

KEY RESULTS

In total, 47.3% had at least one MH condition. Compared to those without MH conditions, Veterans with MH conditions had less fragmented care, with fewer non-MH providers (IRR = 0.96; 95% CI: 0.96-0.96) and more concentrated care with their usual provider (OR = 1.08 for a higher UPC; 95% CI: 1.07, 1.09) in adjusted models. Secondary analyses showed that those with individual MH conditions (e.g., depression) had fewer non-MH providers (IRR range: 0.86-0.98) and more concentrated care (OR range: 1.04-1.20). A similar pattern was observed when examining combinations of MH conditions (IRR range: 0.80-0.90; OR range: 1.16-1.30).

CONCLUSIONS

Contrary to expectations, having a MH condition was associated with less fragmented non-MH care among older, high-risk Veterans. Further research will determine if this is due to different needs, underuse, or appropriate use of healthcare.

Racial/Ethnic Variation in Veterans Health Administration COVID-19 Vaccine Uptake

INTRODUCTION

Equitable COVID-19 vaccine access is imperative to mitigating negative COVID-19 impacts among racial/ethnic minorities. U.S. racial/ethnic minorities have lower COVID-19 vaccination rates than Whites despite higher COVID-19 death/case rates. The Veterans Health Administration provides the unique context of a managed care system with few access barriers. This study evaluates race/ethnicity as a predictor of Veterans Health Administration COVID-19 vaccination.

METHODS

The cohort was composed of Veterans Health Administration outpatient users aged ≥65 years (N=3,474,874). COVID-19 vaccination was assessed between December 14, 2020 and February 23, 2021. Multivariable logistic regressions were conducted, controlling for demographics, medical comorbidity, and influenza vaccination history. Proximity to Indian Health Service Contract Health Service Delivery Areas was tested as a moderator. Data analyses were conducted during 2021.

RESULTS

Blacks (OR=1.28, 95% CI=1.17, 1.40), Hispanics (OR=1.15, 95% CI=1.05, 1.25), and Asians (OR=1.21, 95% CI=1.02, 1.43) were more likely than Whites to receive Veterans Health Administration COVID-19 vaccinations. American Indian/Alaska Natives were less likely than Whites to receive Veterans Health Administration COVID-19 vaccinations, but only those residing in Contract Health Service Delivery Area counties (OR= 0.58, 95% CI= 0.47, 0.72). Influenza vaccine history positively predicted COVID-19 vaccine uptake (OR= 2.28, 95% CI=2.22, 2.34).

CONCLUSIONS

In the Veterans Health Administration, compared with the general U.S. population, COVID-19 vaccine receipt is higher among most racial/ethnic minority groups than Whites, suggesting reduced vaccination barriers . The Indian Health Service may provide a safety net for American Indian/Alaska Native populations. Addressing vaccination access barriers in non-Veterans Health Administration settings can potentially reduce racial/ethnic disparities.

Peeking into a black box, the fairness and generalizability of a MIMIC-III benchmarking model

As artificial intelligence (AI) makes continuous progress to improve quality of care for some patients by leveraging ever increasing amounts of digital health data, others are left behind. Empirical evaluation studies are required to keep biased AI models from reinforcing systemic health disparities faced by minority populations through dangerous feedback loops. The aim of this study is to raise broad awareness of the pervasive challenges around bias and fairness in risk prediction models. We performed a case study on a MIMIC-trained benchmarking model using a broadly applicable fairness and generalizability assessment framework. While open-science benchmarks are crucial to overcome many study limitations today, this case study revealed a strong class imbalance problem as well as fairness concerns for Black and publicly insured ICU patients. Therefore, we advocate for the widespread use of comprehensive fairness and performance assessment frameworks to effectively monitor and validate benchmark pipelines built on open data resources.

Using an Automated Electronic Health Record Score To Estimate Life Expectancy In Men Diagnosed With Prostate Cancer In The Veterans Health Administration

OBJECTIVES

To determine if an automatically calculated electronic health record score can estimate intermediate-term life expectancy in men with prostate cancer to provide guideline concordant care.

METHODS

We identified all men (n = 36,591) diagnosed with prostate cancer in 2013-2015 in the VHA. Of the 36,591, 35,364 (96.6%) had an available Care Assessment Needs (CAN) score (range: 0-99) automatically calculated in the 30 days prior to the date of diagnosis. It was designed to estimate short-term risks of hospitalization and mortality. We fit unadjusted and multivariable Cox proportional hazards regression models to determine the association between the CAN score and overall survival among men with prostate cancer. We compared CAN score performance to two established comorbidity measures: The Charlson Comorbidity Index and Prostate Cancer Comorbidity Index (PCCI).

RESULTS

Among 35,364 men, the CAN score correlated with overall stage, with mean scores of 46.5 ( ± 22.4), 58.0 ( ± 24.4), and 68.1 ( ± 24.3) in localized, locally advanced, and metastatic disease, respectively. In both unadjusted and adjusted models for prostate cancer risk, the CAN score was independently associated with survival (HR = 1.23 95%CI 1.22-1.24 & adjusted HR = 1.17 95%CI 1.16-1.18 per 5-unit change, respectively). The CAN score (overall C-Index 0.74) yielded better discrimination (AUC = 0.76) than PCCI (AUC = 0.65) or Charlson Comorbidity Index (AUC = 0.66) for 5-year survival.

CONCLUSION

The CAN score is strongly associated with intermediate-term survival following a prostate cancer diagnosis. The CAN score is an example of how learning health care systems can implement multi-dimensional tools to provide fully automated life expectancy estimates to facilitate patient-centered cancer care.

Psychometric evaluation of a patient-reported item bank for healthcare engagement

PURPOSE

Healthcare engagement is a core measurement target for efforts to improve healthcare systems. This construct is broadly defined as the extent to which healthcare services represent collaborative partnerships with patients. Previous qualitative work operationalized healthcare engagement as generalized self-efficacy in four related subdomains: self-management, collaborative communication, health information use, and healthcare navigation. Building on this work, our objective was to establish a healthcare engagement instrument that is sufficiently unidimensional to yield a single score.

METHOD

We conducted cognitive interviews followed by a nation-wide mail survey of US Veteran Administration (VA) healthcare users. Data were collected on 49 candidate healthcare engagement items, as well as measures of self-efficacy for managing symptoms, provider communication, and perceived access. Items were subjected to exploratory bifactor, statistical learning, and IRT analyses.

RESULTS

Cognitive interviews were completed by 56 patients and 9552 VA healthcare users with chronic conditions completed the mail survey. Participants were mostly white and male but with sizable minority participation. Psychometric analyses and content considerations reduced the item pool to 23 items, which demonstrated a strong general factor (OmegaH of .89). IRT analyses revealed a high level of reliability across the trait range and little DIF across groups. Most health information use items were removed during analyses, suggesting a more independent role for this domain.

CONCLUSION

We provide quantitative evidence for a relatively unidimensional measure of healthcare engagement. Despite developed with VA healthcare users, the measure is intended for general use. Future work includes short-form development and validation with other patient groups.

Demystifying machine learning: a primer for physicians

Machine learning is a tool for analysing digitised data sets and formulating predictions that can optimise clinical decision-making. It aims to identify complex patterns in large data sets and encode them into models that can then classify new unseen cases or make predictions on new data. Machine learning methods take several forms and individual models can be of many different types. More than 50 models have been approved for use in routine healthcare, and the numbers continue to grow exponentially. The reliability and robustness of any model depends on multiple factors, including the quality and quantity of the data used to develop the models, and the selection of features in the data considered most important to maximising accuracy. In ensuring models are safe, effective and reproducible in routine care, physicians need to have some understanding of how these models are developed and evaluated, and to collaborate with data and computer scientists in their design and validation. This narrative review introduces principles, methods and examples of machine learning in a way that does not require mastery of highly complex statistical and computational concepts.

Automated EHR score to predict COVID-19 outcomes at US Department of Veterans Affairs

The sudden emergence of COVID-19 has brought significant challenges to the care of Veterans. An improved ability to predict a patient's clinical course would facilitate optimal care decisions, resource allocation, family counseling, and strategies for safely easing distancing restrictions. The Care Assessment Need (CAN) score is an existing risk assessment tool within the Veterans Health Administration (VA), and produces a score from 0 to 99, with a higher score correlating to a greater risk. The model was originally designed for the nonacute outpatient setting and is automatically calculated from structured data variables in the electronic health record. This multisite retrospective study of 6591 Veterans diagnosed with COVID-19 from March 2, 2020 to May 26, 2020 was designed to assess the utility of repurposing the CAN score as objective and automated risk assessment tool to promptly enhance clinical decision making for Veterans diagnosed with COVID-19. We performed bivariate analyses on the dichotomized CAN 1-year mortality score (high vs. low risk) and each patient outcome using Chi-square tests of independence. Logistic regression models using the continuous CAN score were fit to assess its predictive power for outcomes of interest. Results demonstrated that a CAN score greater than 50 was significantly associated with the following outcomes after positive COVID-19 test: hospital admission (OR 4.6), prolonged hospital stay (OR 4.5), ICU admission (3.1), prolonged ICU stay (OR 2.9), mechanical ventilation (OR 2.6), and mortality (OR 7.2). Repurposing the CAN score offers an efficient way to risk-stratify COVID-19 Veterans. As a result of the compelling statistical results, and automation, this tool is well positioned for broad use across the VA to enhance clinical decision-making.