Accuracy of an Electronic Health Record Patient Linkage Module Evaluated between Neighboring Academic Health Care Centers

A qualitative study of primary care clinician's approach to ending cervical cancer screening in older women in the United States

The United States Preventive Services Task Force (USPSTF) recommends that cervical cancer screening end in average-risk patients with a cervix at 65 years of age if adequate screening measures have been met, defined as having 1) at least three normal consecutive cytology (Pap) tests, or 2) two normal cytology tests and/or two negative high-risk human papillomavirus tests between ages 55-65; the last test should be performed within the prior 5 years. Up to 60 % of all women aged 65 years and older who are ending screening do not meet the criteria for adequate screening. The objective of this study was to understand the process and approach that healthcare clinicians use to determine eligibility to end cervical cancer screening. In 2021 we conducted semi-structured interviews in San Francisco, CA with twelve healthcare clinicians: two family medicine physicians, three general internal medicine physicians, two obstetrician/gynecologists and five nurse practitioners. Thematic analysis, using inductive and deductive coding, was utilized. Three major themes emerged: following guidelines, relying on self-reported data regarding prior screening, and considering sexual activity as a factor in the decision to end screening. All interviewees endorsed following the USPSTF guidelines and they utilized self-report to determine eligibility to end screening. Clinicians' approach was dependent in part on their judgement about the reliability of the patient to convey their screening history. Sexual activity of the patient was considered when making clinical recommendations. Shared decision-making was often utilized. Clinicians voiced a strong reliance on self-reported screening history to end cervical cancer screening.

Engaging Patients and Other Stakeholders in "Designing for Dissemination" of Record Linkage Methods and Tools

BACKGROUND

Novel record linkage (RL) methods have the potential to enhance clinical informatics by integrating patient data from multiple sources-including electronic health records, insurance claims, and digital health devices-to inform patient-centered care. Engaging patients and other stakeholders in the use of RL methods in patient-centered outcomes research (PCOR) is a key step in ensuring RL methods are viewed as acceptable, appropriate, and useful. The University of Colorado Record Linkage (CURL) platform empowers the use of RL in PCOR.

OBJECTIVES

This study aimed to describe the process of engaging patients and other stakeholders in the design of an RL dissemination package to support the use of RL methods in PCOR.

METHODS

Customer discovery, value proposition design, and user experience methods were used to iteratively develop an RL dissemination package that includes animated explainer videos for patients and an RL research planning workbook for researchers. Patients and other stakeholders (researchers, data managers, and regulatory officials) were engaged in the RL dissemination package design.

RESULTS

Patient partners emphasized the importance of conveying how RL methods may benefit patients and the rules researchers must follow to protect the privacy and security of patient data. Other stakeholders described accuracy, flexibility, efficiency, and data security compared with other available RL solutions. Dissemination package communication products reflect the value propositions identified by key stakeholders. As prioritized by patients, the animated explainer videos emphasize the data privacy and security processes and procedures employed when performing research using RL. The RL workbook addresses researchers' and data managers' needs to iteratively design RL projects and provides accompanying resources to alleviate leadership and regulatory officials' concerns about data regulation compliance.

CONCLUSION

Dissemination products to promote adoption and use of CURL include materials to facilitate patient engagement in RL research and investigator step-by-step decision-making materials about the integration of RL methods in PCOR.

Optimizing Patient Record Linkage in a Master Patient Index Using Machine Learning: Algorithm Development and Validation

BACKGROUND

To provide quality care, modern health care systems must match and link data about the same patient from multiple sources, a function often served by master patient index (MPI) software. Record linkage in the MPI is typically performed manually by health care providers, guided by automated matching algorithms. These matching algorithms must be configured in advance, such as by setting the weights of patient attributes, usually by someone with knowledge of both the matching algorithm and the patient population being served.

OBJECTIVE

We aimed to develop and evaluate a machine learning-based software tool, which automatically configures a patient matching algorithm by learning from pairs of patient records previously linked by humans already present in the database.

METHODS

We built a free and open-source software tool to optimize record linkage algorithm parameters based on historical record linkages. The tool uses Bayesian optimization to identify the set of configuration parameters that lead to optimal matching performance in a given patient population, by learning from prior record linkages by humans. The tool is written assuming only the existence of a minimal HTTP application programming interface (API), and so is agnostic to the choice of MPI software, record linkage algorithm, and patient population. As a proof of concept, we integrated our tool with SantéMPI, an open-source MPI. We validated the tool using several synthetic patient populations in SantéMPI by comparing the performance of the optimized configuration in held-out data to SantéMPI's default matching configuration using sensitivity and specificity.

RESULTS

The machine learning-optimized configurations correctly detect over 90% of true record linkages as definite matches in all data sets, with 100% specificity and positive predictive value in all data sets, whereas the baseline detects none. In the largest data set examined, the baseline matching configuration detects possible record linkages with a sensitivity of 90.2% (95% CI 88.4%-92.0%) and specificity of 100%. By comparison, the machine learning-optimized matching configuration attains a sensitivity of 100%, with a decreased specificity of 95.9% (95% CI 95.9%-96.0%). We report significant gains in sensitivity in all data sets examined, at the cost of only marginally decreased specificity. The configuration optimization tool, data, and data set generator have been made freely available.

CONCLUSIONS

Our machine learning software tool can be used to significantly improve the performance of existing record linkage algorithms, without knowledge of the algorithm being used or specific details of the patient population being served.

How Image Exchange Breaks Down: the Image Library Perspective

While in the last decade there has been significant technical infrastructure development to support standards-based image exchange through organizations like Integrating the Healthcare Enterprise, Carequality, DICOM, and HL7 FHIR, the human operationalization of such infrastructure using centralized, intuitive, standards-based applications remains the cornerstone of effective and reliable electronic image exchange. Image libraries managing the highly transactional and often uncertain inflows and outflows of images have a unique perspective on the challenges of image exchange. This manuscript will summarize frequent collaboration and communication, release of information, staffing, technology, information localization, and analytics difficulties for image exchange from the perspective of the image library staff managing the transactions.

Community Income, Healthy Food Access, and Repeat Surgery for Kidney Stones

OBJECTIVE

To determine if limited food access census tracts and food swamp census tracts are associated with increased risk for repeat kidney stone surgery. And to elucidate the relationship between community-level food retail environment relative to community-level income on repeat stone surgery over time.

METHODS

Data were abstracted from the University of California, San Francisco Information Commons. Adult patients were included if they underwent at least one urologic stone procedure. Census tracts from available geographical data were mapped using Food Access Research Atlas data from the United States Department of Agriculture Economic Research Service. Kaplan-Meier curves were employed to illustrate time to a second surgical procedure over 5 years, and log-rank tests were used to test for statistically significant differences. A multivariate Cox regression model was used to generate hazard ratios for undergoing second surgery by group.

RESULTS

A total of 1496 patients were included in this analysis. Repeat stone surgery occurred in 324 patients. Kaplan-Meier curves demonstrated a statistically significant difference in curves depicting patients living in low income census tracts (LICTs) vs those not living in LICTs (P <.001). On Cox regression models, patients in LICTs had significantly higher risk of undergoing repeat surgery (P = .011). Patients from limited food access census tracts and food swamp census tracts did not have a significantly higher adjusted risk of undergoing second surgery (P = .11 and P = .88, respectively).

CONCLUSION

Income more so than food access associates with increased risk of repeat kidney stone surgery. Further research is needed to explore the interaction between low socioeconomic status and kidney stone outcomes.

The Cosmos Collaborative: A Vendor-Facilitated Electronic Health Record Data Aggregation Platform

Objective

Learning healthcare systems use routinely collected data to generate new evidence that informs future practice. While implementing an electronic health record (EHR) system can facilitate this goal for individual institutions, meaningfully aggregating data from multiple institutions can be more empowering. Cosmos is a cross-institution, single EHR vendor-facilitated data aggregation tool. This work aims to describe the initiative and illustrate its potential utility through several use cases.

Methods

Cosmos is designed to scale rapidly by leveraging preexisting agreements, clinical health information exchange networks, and data standards. Data are stored centrally as a limited dataset, but the customer facing query tool limits results to prevent patient reidentification.

Results

In 2 years, Cosmos grew to contain EHR data of more than 60 million patients. We present practical examples illustrating how Cosmos could further efforts in chronic disease surveillance (asthma and obesity), syndromic surveillance (seasonal influenza and the 2019 novel coronavirus), immunization adherence and adverse event reporting (human papilloma virus and measles, mumps, rubella, and varicella vaccination), and health services research (antibiotic usage for upper respiratory infection).

Discussion

A low barrier of entry for Cosmos allows for the rapid accumulation of multi-institutional and mostly de-duplicated EHR data to power research and quality improvement queries characteristic of learning healthcare systems. Limitations are being vendor-specific, an “all or none” contribution model, and the lack of control over queries run on an institution’s healthcare data.

Conclusion

Cosmos provides a model for within-vendor data standardization and aggregation and a steppingstone for broader intervendor interoperability.

Alcohol use disorder treatment and outcomes among hospitalized adults with alcoholic hepatitis

Purpose

: The burden of alcohol-associated liver disease (ALD) in the United States (US) has continued to worsen in the background of rising rates of alcohol use disorder. Patients with ALD present to care at a late stage, often with the sequela of liver decompensation, such as gastrointestinal bleeding and infection. ALD is now the leading indication for liver transplantation. We aimed to measure the quality of care delivered to hospitalized patients with alcoholic hepatitis (AH) across 3 domains: 1) alcohol-use disorder (AUD) care, 2) inpatient cirrhosis care, and 3) alcohol-associated liver disease (ALD) care-and observe associations between quality of care and outcomes.

Methods

: We included hospital encounters between January 1, 2016 and January 1, 2019 to a large, diverse integrated health system for AH with active alcohol use within the prior 60 days. The diagnosis of AH was determined based on previously published clinical and laboratory criteria. Quality indicator (QI) pass rates were calculated as the proportion of patients eligible for each indicator who received the QI within the timeframe specified. We then evaluated the association between the receipt of all QIs and 6-month mortality, as well as AUD-specific QIs and 30-day readmission.

Results

: Of the 179 patients, the median age was 47 years-old, 59.2% were male and 49.2% were non-Hispanic White. The median Model for End-Stage Liver Disease-Sodium score was 25, while the median discriminant function was 33. Patients were followed for an average of 21 months. Overall, 14% of patients died during the index hospitalization while 17.3% died following discharge and 24.8% were re-admitted within 30-days. QI pass-rates were variable across the different domains. Few patients received AUD care-pass rates for receipt of pharmacotherapy and behavioral therapy at 6 months were only 19.1% and 35.1%, respectively. There was a significant association between receiving behavioral therapy and 6-month mortality-3% vs 18%, p = 0.05.

Conclusion

: The quality of care received during hospital encounters for AH is variable, and AUD-specific therapy is low. Future quality of care initiatives are warranted to link patients to AUD treatment to ensure optimal care and maximize patients survival in this at-risk population.