Extracting patient-level data from the electronic health record: Expanding opportunities for health system research.

BACKGROUND: Epidemiological studies of interstitial lung disease (ILD) are limited by small numbers and tertiary care bias. Investigators have leveraged the widespread use of electronic health records (EHRs) to overcome these limitations, but struggle to extract patient-level, longitudinal clinical data needed to address many important research questions. We hypothesized that we could automate longitudinal ILD cohort development using the EHR of a large, community-based healthcare system. STUDY DESIGN AND METHODS: We applied a previously validated algorithm to the EHR of a community-based healthcare system to identify ILD cases between 2012-2020. We then extracted disease-specific characteristics and outcomes using fully automated data-extraction algorithms and natural language processing of selected free-text. RESULTS: We identified a community cohort of 5,399 ILD patients (prevalence = 118 per 100,000). Pulmonary function tests (71%) and serologies (54%) were commonly used in the diagnostic evaluation, whereas lung biopsy was rare (5%). IPF was the most common ILD diagnosis (n = 972, 18%). Prednisone was the most commonly prescribed medication (911, 17%). Nintedanib and pirfenidone were rarely prescribed (n = 305, 5%). ILD patients were high-utilizers of inpatient (40%/year hospitalized) and outpatient care (80%/year with pulmonary visit), with sustained utilization throughout the post-diagnosis study period. DISCUSSION: We demonstrated the feasibility of robustly characterizing a variety of patient-level utilization and health services outcomes in a community-based EHR cohort. This represents a substantial methodological improvement by alleviating traditional constraints on the accuracy and clinical resolution of such ILD cohorts; we believe this approach will make community-based ILD research more efficient, effective, and scalable.

Patterns and Factors Associated With Adherence to Lung Cancer Screening in Diverse Practice Settings.

Importance: For lung cancer screening to confer mortality benefit, adherence to annual screening with low-dose computed tomography scans is essential. Although the National Lung Screening Trial had an adherence rate of 95%, current data are limited on screening adherence across diverse practice settings in the United States. Objective: To evaluate patterns and factors associated with adherence to annual screening for lung cancer after negative results of a baseline examination, particularly in centralized vs decentralized screening programs. Design, Setting, and Participants: This observational cohort study was conducted at 5 academic and community-based sites in North Carolina and California among 2283 individuals screened for lung cancer between July 1, 2014, and March 31, 2018, who met US Preventive Services Task Force eligibility criteria, had negative results of a baseline screening examination (American College of Radiology Lung Imaging Reporting and Data System category 1 or 2), and were eligible to return for a screening examination in 12 months. Exposures: To identify factors associated with adherence, the association of adherence with selected baseline demographic and clinical characteristics, including type of screening program, was estimated using multivariable logistic regression. Screening program type was classified as centralized if individuals were referred through a lung cancer screening clinic or program and as decentralized if individuals had a direct clinician referral for the baseline low-dose computed tomography scan. Main Outcomes and Measures: Adherence to annual lung cancer screening, defined as a second low-dose computed tomography scan within 11 to 15 months after baseline screening. Results: Among the 2283 eligible individuals (1294 men [56.7%]; mean [SD] age, 64.9 [5.8] years; 1160 [50.8%] aged ≥65 years) who had negative screening results at baseline, overall adherence was 40.2% (n = 917), with higher adherence among those who underwent screening through centralized (46.0% [478 of 1039]) vs decentralized (35.3% [439 of 1244]) programs. The independent factor most strongly associated with adherence was type of screening program, with a 2.8-fold increased likelihood of adherence associated with centralized screening (adjusted odds ratio [aOR], 2.78; 95% CI, 1.99-3.88). Another associated factor was age (65-69 vs 55-59 years: aOR, 1.38; 95% CI, 1.07-1.77; 70-74 vs 55-59 years: aOR, 1.47; 95% CI, 1.10-1.96). Conclusions and Relevance: After negative results of a baseline examination, adherence to annual lung cancer screening was suboptimal, although adherence was higher among individuals who were screened through a centralized program. These results support the value of centralized screening programs and the need to further implement strategies that improve adherence to annual screening for lung cancer.

Impact of Idiopathic Pulmonary Fibrosis on Longitudinal Health-care Utilization in a Community-Based Cohort of Patients.

BACKGROUND: Idiopathic pulmonary fibrosis (IPF) is a rare, chronic lung disease associated with substantial symptom burden, morbidity, and cost. Delivery of high-quality effective care in IPF requires understanding health-care resource utilization (HRU) patterns; however, longitudinal data from real-world populations are limited. RESEARCH QUESTION: This study aimed to define HRU attributable to IPF by evaluating a longitudinal cohort of community patients with IPF compared with matched control subjects. STUDY DESIGN AND METHODS: Incident IPF cases were identified in the Kaiser Permanente Northern California electronic health records (2000-2015) using case-validated code-based algorithms. IPF cases were compared with matched control subjects by age, sex, and length of enrollment. Annual rates of HRU measures were assessed during the 5 years pre- and postdiagnosis. Poisson generalized estimating equations were used to estimate adjusted case-control differences in HRU. IPF treatment trends were assessed before and after the availability of IPF-specific medications. RESULTS: A total of 691 IPF cases were identified and matched with 3,452 control subjects. Adjusted rates of all diagnostic procedures were significantly increased (P < .001) for IPF cases compared with control subjects in both the pre- and postindex periods, including chest CT scans (pre-relative risk [RR], 80.35; post-RR, 32.79), 6-min walk tests (pre-RR, 20.81; post-RR, 34.49), and pulmonary function tests (pre-RR, 9.50; post-RR, 13.24). All-cause hospitalizations (pre-RR, 1.42; post-RR, 2.33) and outpatient visits (pre-RR, 1.22; post-RR, 1.80) were significantly higher among cases compared with control subjects during both the preindex (P < .05) and postindex (P < .001) periods. We observed use of immunosuppressive and IPF-specific therapies prior to diagnosis, and high rates of corticosteroid use before and after diagnosis. INTERPRETATION: This study defines a marked increase in HRU in patients with IPF compared with control subjects, with accelerated use beginning at least 1 year prediagnosis and elevated use sustained over the following 5 years. To our knowledge, this is the first study to evaluate longitudinal medication trends in IPF. Collectively, this information is foundational to advancing IPF care delivery models and supporting clinical decision-making.

Effectiveness of a Patient Education Class to Enhance Knowledge about Lung Cancer Screening: a Quality Improvement Evaluation.

Best practices to facilitate high-quality shared decision-making for lung cancer screening (LCS) are not well established. In our LCS program, patients are first referred to attend a free group education class on LCS, taught by designated clinician specialists, before a personal shared decision-making visit is scheduled. We conducted an evaluation on  the effectiveness of this class to enhance patient knowledge and shared decision-making about LCS. For quality improvement purposes, participants were asked to complete one-page surveys immediately before and after class to assess knowledge and decision-making capacity regarding LCS. To evaluate knowledge gained, we tabulated the distributions of correct, incorrect, unsure, and missing responses to eight true-false statements included on both pre- and post-class surveys and assessed pre-post differences in the number of correct responses. To evaluate decision-making capacity, we tabulated the distributions of post-class responses to items on decision uncertainty. From June 2017 to August 2018, 680 participants completed both pre- and post-class surveys. Participants had generally poor baseline knowledge about LCS. The proportion who responded correctly to each knowledge-related statement increased pre- to post-class, with a mean difference of 0.9 (paired t test, p < 0.0001) in the total number of correct responses between surveys. About 70% reported having all the information needed to make a screening decision. Our results suggest that a well-designed group education class is an effective system-level approach for initially educating and equipping patients with appropriate knowledge to make informed decisions about LCS.

Monitoring Lung Cancer Screening Use and Outcomes at Four Cancer Research Network Sites.

RATIONALE: Lung cancer screening registries can monitor screening outcomes and improve quality of care. OBJECTIVES: To describe nascent lung cancer screening programs and share efficient data collection approaches for mandatory registry reporting in four integrated health care systems of the National Cancer Institute-funded Cancer Research Network. METHODS: We documented the distinctive characteristics of lung cancer screening programs, and we provide examples of strategies to facilitate data collection and describe early challenges and possible solutions. In addition, we report preliminary data on use and outcomes of screening with low-dose computed tomography at each of the participating sites. RESULTS: Programs varied in approaches to confirming patient eligibility, ordering screening low-dose computed tomographic scans, and coordinating follow-up care. Most data elements were collected from structured fields in electronic health records, but sites also made use of standardized order templates, local procedure codes, identifiable hashtags in radiology reports, and natural language processing algorithms. Common challenges included incomplete documentation of tobacco smoking history, difficulty distinguishing between scans performed for screening versus diagnosis or surveillance, and variable adherence with use of standardized templates. Adherence with eligibility criteria as well as the accuracy and completeness of data collection appeared to depend at least partly on availability of personnel and other resources to support the successful implementation of screening. CONCLUSIONS: To maximize the effectiveness of lung cancer screening, minimize the burden of data collection, and facilitate research and quality improvement, clinical workflow and information technology should be purposefully designed to ensure that patients meet eligibility criteria and receive appropriate follow-up testing.

Overexpression of Sprouty 2 in mouse lung epithelium inhibits urethane-induced tumorigenesis.

Members of the Sprouty family encode novel proteins that are thought to function primarily as intracellular antagonists of the Ras-signaling pathway. Increased Ras signaling is a critical characteristic of human lung adenocarcinoma, the most common type of non-small cell lung cancer. Sprouty 2 is expressed in the lung epithelium, the tissue layer from which lung cancers arise. We hypothesized that overexpression of Sprouty 2 in the distal lung epithelium would inhibit lung tumorigenesis. To test the hypothesis, the consequences of overexpressing Sprouty 2 in the distal lung epithelium on urethane-induced mouse lung tumorigenesis were determined. Urethane is a chemical carcinogen found in tobacco smoke that causes activating mutations in Kras and induces lung tumors in mice. Sprouty 2-overexpressor mice developed significantly fewer lung tumors compared with their littermate controls (13.2 +/- 1.1 versus 18.1 +/- 1.3, P = 0.006). Tumor diameter was also significantly smaller in Sprouty 2 overexpressors (0.85 mm +/- 0.03 versus 0.95 mm +/- 0.02, P = 0.005). Sprouty 2 overexpression did not alter Kras mutational frequencies in urethane-induced tumors, suggesting that the tumor-suppressing effect of Sprouty 2 overexpression acts at a stage after Kras mutation, perhaps by interfering with receptor tyrosine kinase-induced signaling. These results demonstrate that Sprouty 2 overexpression inhibited both tumor initiation and subsequent tumor growth.

An FGF response pathway that mediates hepatic gene induction in embryonic endoderm cells.

While particular combinations of mesodermal signals are known to induce distinct tissue-specific programs in the endoderm, there is little information about the response pathways within endoderm cells that control their specification. We have used signaling inhibitors on embryo tissue explants and whole-embryo cultures as well as genetic approaches to reveal part of an intracellular network by which FGF signaling helps induce hepatic genes and stabilize nascent hepatic cells within the endodermal epithelium. Specifically, we found that hepatic gene induction is elicited by an FGF/MAPK pathway. Although the PI3K pathway is activated in foregut endoderm cells, its inhibition does not block hepatic gene induction in explants; however, it does block tissue growth. We also found that at the onset of hepatogenesis, the FGF/MAPK and PI3K pathways do not crossregulate in the endoderm. The finding of separate pathways for endoderm tissue specification and growth provides insights for guiding cellular regeneration and stem cell differentiation.

Sprouty genes control diastema tooth development via bidirectional antagonism of epithelial-mesenchymal FGF signaling.

Unlike humans, who have a continuous row of teeth, mice have only molars and incisors separated by a toothless region called a diastema. Although tooth buds form in the embryonic diastema, they regress and do not develop into teeth. Here, we identify members of the Sprouty (Spry) family, which encode negative feedback regulators of fibroblast growth factor (FGF) and other receptor tyrosine kinase signaling, as genes that repress diastema tooth development. We show that different Sprouty genes are deployed in different tissue compartments--Spry2 in epithelium and Spry4 in mesenchyme--to prevent diastema tooth formation. We provide genetic evidence that they function to ensure that diastema tooth buds are refractory to signaling via FGF ligands that are present in the region and thus prevent these buds from engaging in the FGF-mediated bidirectional signaling between epithelium and mesenchyme that normally sustains tooth development.

Increasing Fgf4 expression in the mouse limb bud causes polysyndactyly and rescues the skeletal defects that result from loss of Fgf8 function.

A major function of the limb bud apical ectodermal ridge (AER) is to produce fibroblast growth factors (FGFs) that signal to the underlying mesenchyme. Previous studies have suggested that of the four FGF genes specifically expressed in the mouse AER, Fgf8 is unique not only in its expression pattern, but also because it is the only such FGF gene that causes limb skeletal abnormalities when individually inactivated. However, when both Fgf8 and Fgf4 are simultaneously inactivated in the AER, the limb does not develop. One possible explanation for these observations is that although both of these FGF family members contribute to limb development, Fgf8 has functions that Fgf4 cannot perform. To test this hypothesis, we used a novel method to substitute Fgf4 for Fgf8 expression in the developing limb bud by concomitantly activating a conditional Fgf4 gain-of-function allele and inactivating an Fgf8 loss-of-function allele in the same cells via Cre-mediated recombination. Our data show that when Fgf4 is expressed in place of Fgf8, all of the skeletal defects caused by inactivation of Fgf8 are rescued, conclusively demonstrating that FGF4 can functionally replace FGF8 in limb skeletal development. We also show that the increase in FGF signaling that occurs when the Fgf4 gain-of-function allele is activated in a wild-type limb bud causes formation of a supernumerary posterior digit (postaxial polydactyly), as well as cutaneous syndactyly between all the digits. These data underscore the importance of controlling the level of FGF gene expression for normal limb development.

Sprouty2, a mouse deafness gene, regulates cell fate decisions in the auditory sensory epithelium by antagonizing FGF signaling.

The auditory sensory epithelium (organ of Corti), where sound waves are converted to electrical signals, comprises a highly ordered array of sensory receptor (hair) cells and nonsensory supporting cells. Here, we report that Sprouty2, which encodes a negative regulator of signaling via receptor tyrosine kinases, is required for normal hearing in mice, and that lack of SPRY2 results in dramatic perturbations in organ of Corti cytoarchitecture: instead of two pillar cells, there are three, resulting in the formation of an ectopic tunnel of Corti. We demonstrate that these effects are due to a postnatal cell fate transformation of a Deiters' cell into a pillar cell. Both this cell fate change and hearing loss can be partially rescued by reducing Fgf8 gene dosage in Spry2 null mutant mice. Our results provide evidence that antagonism of FGF signaling by SPRY2 is essential for establishing the cytoarchitecture of the organ of Corti and for hearing.

Expression of mutant human epidermal receptor 3 attenuates lung fibrosis and improves survival in mice.

Neuregulin-1 (NRG-1), binding to the human epidermal growth factor receptor HER2/HER3, plays a role in pulmonary epithelial cell proliferation and recovery from injury in vitro. We hypothesized that activation of HER2/HER3 by NRG-1 would also play a role in recovery from in vivo lung injury. We tested this hypothesis using bleomycin lung injury of transgenic mice incapable of signaling through HER2/HER3 due to lung-specific dominant-negative HER3 (DNHER3) expression. In animals expressing DNHER3, protein leak, cell infiltration, and NRG-1 levels in bronchoalveolar lavage fluid increased after injury, similar to that in nontransgenic littermate control animals. However, HER2/HER3 was not activated, and DNHER3 animals displayed fewer lung morphological changes at 10 and 21 days after injury (P = 0.01). In addition, they contained 51% less collagen in injured lungs (P = 0.04). Transforming growth factor-beta1 did not increase in bronchoalveolar lavage fluid from DNHER3 mice compared with nontransgenic littermate mice (P = 0.001), suggesting that a mechanism for the decreased fibrosis was lack of transforming growth factor-beta1 induction in DNHER3 mice. Severe lung injury (0.08 units bleomycin) resulted in 80% mortality of nontransgenic mice, but only 35% mortality of DNHER3 transgenic mice (P = 0.04). Thus inhibition of HER2/HER3 signaling protects against pulmonary fibrosis and improves survival.