Performance of an automated deep learning algorithm to identify hepatic steatosis within noncontrast computed tomography scans among people with and without HIV.

PURPOSE: Hepatic steatosis (fatty liver disease) affects 25% of the world's population, particularly people with HIV (PWH). Pharmacoepidemiologic studies to identify medications associated with steatosis have not been conducted because methods to evaluate liver fat within digitized images have not been developed. We determined the accuracy of a deep learning algorithm (automatic liver attenuation region-of-interest-based measurement [ALARM]) to identify steatosis within clinically obtained noncontrast abdominal CT images compared to manual radiologist review and evaluated its performance by HIV status. METHODS: We performed a cross-sectional study to evaluate the performance of ALARM within noncontrast abdominal CT images from a sample of patients with and without HIV in the US Veterans Health Administration. We evaluated the ability of ALARM to identify moderate-to-severe hepatic steatosis, defined by mean absolute liver attenuation <40 Hounsfield units (HU), compared to manual radiologist assessment. RESULTS: Among 120 patients (51 PWH) who underwent noncontrast abdominal CT, moderate-to-severe hepatic steatosis was identified in 15 (12.5%) persons via ALARM and 12 (10%) by radiologist assessment. Percent agreement between ALARM and radiologist assessment of absolute liver attenuation <40 HU was 95.8%. Sensitivity, specificity, positive predictive value, and negative predictive value of ALARM were 91.7% (95%CI, 51.5%-99.8%), 96.3% (95%CI, 90.8%-99.0%), 73.3% (95%CI, 44.9%-92.2%), and 99.0% (95%CI, 94.8%-100%), respectively. No differences in performance were observed by HIV status. CONCLUSIONS: ALARM demonstrated excellent accuracy for moderate-to-severe hepatic steatosis regardless of HIV status. Application of ALARM to radiographic repositories could facilitate real-world studies to evaluate medications associated with steatosis and assess differences by HIV status.

Cancer tracking system improves timeliness of liver cancer care at a Veterans Hospital: A comparison of cohorts before and after implementation of an automated care coordination tool.

INTRODUCTION: Hepatocellular carcinoma (HCC) requires complex care coordination. Patient safety may be compromised with untimely follow-up of abnormal liver imaging. This study evaluated whether an electronic case-finding and tracking system improved timeliness of HCC care. METHODS: An electronic medical record-linked abnormal imaging identification and tracking system was implemented at a Veterans Affairs Hospital. This system reviews all liver radiology reports, generates a queue of abnormal cases for review, and maintains a queue of cancer care events with due dates and automated reminders. This is a pre-/post-intervention cohort study to evaluate whether implementation of this tracking system reduced time between HCC diagnosis and treatment and time between first liver image suspicious for HCC, specialty care, diagnosis, and treatment at a Veterans Hospital. Patients diagnosed with HCC in the 37 months before tracking system implementation were compared to patients diagnosed with HCC in the 71 months after its implementation. Linear regression was used to calculate mean change in relevant intervals of care adjusted for age, race, ethnicity, BCLC stage, and indication for first suspicious image. RESULTS: There were 60 patients pre-intervention and 127 post-intervention. In the post-intervention group, adjusted mean time from diagnosis to treatment was 36 days shorter (p = 0.007), time from imaging to diagnosis 51 days shorter (p = 0.21), and time from imaging to treatment 87 days shorter (p = 0.05). Patients whose imaging was performed for HCC screening had the greatest improvement in time from diagnosis to treatment (63 days, p = 0.02) and from first suspicious image to treatment (179 days, p = 0.03). The post-intervention group also had a greater proportion of HCC diagnosed at earlier BCLC stages (p<0.03). CONCLUSIONS: The tracking system improved timeliness of HCC diagnosis and treatment and may be useful for improving HCC care delivery, including in health systems already implementing HCC screening.

Comparison of Natural Language Processing and Manual Coding for the Identification of Cross-Sectional Imaging Reports Suspicious for Lung Cancer.

PURPOSE: To compare the accuracy and reliability of a natural language processing (NLP) algorithm with manual coding by radiologists, and the combination of the two methods, for the identification of patients whose computed tomography (CT) reports raised the concern for lung cancer. METHODS: An NLP algorithm was developed using Clinical Text Analysis and Knowledge Extraction System (cTAKES) with the Yale cTAKES Extensions and trained to differentiate between language indicating benign lesions and lesions concerning for lung cancer. A random sample of 450 chest CT reports performed at Veterans Affairs Connecticut Healthcare System between January 2014 and July 2015 was selected. A reference standard was created by the manual review of reports to determine if the text stated that follow-up was needed for concern for cancer. The NLP algorithm was applied to all reports and compared with case identification using the manual coding by the radiologists. RESULTS: A total of 450 reports representing 428 patients were analyzed. NLP had higher sensitivity and lower specificity than manual coding (77.3% v 51.5% and 72.5% v 82.5%, respectively). NLP and manual coding had similar positive predictive values (88.4% v 88.9%), and NLP had a higher negative predictive value than manual coding (54% v 38.5%). When NLP and manual coding were combined, sensitivity increased to 92.3%, with a decrease in specificity to 62.85%. Combined NLP and manual coding had a positive predictive value of 87.0% and a negative predictive value of 75.2%. CONCLUSION: Our NLP algorithm was more sensitive than manual coding of CT chest reports for the identification of patients who required follow-up for suspicion of lung cancer. The combination of NLP and manual coding is a sensitive way to identify patients who need further workup for lung cancer.

Continuity in a VA patient-centered medical home reduces emergency department visits.

BACKGROUND: One major goal of the Patient-Centered Medical Home (PCMH) is to improve continuity of care between patients and providers and reduce the utilization of non-primary care services like the emergency department (ED). OBJECTIVE: To characterize continuity under the Veterans Health Administration's PCMH model--the Patient Aligned Care Team (PACT), at one large Veterans Affair's (VA's) primary care clinic, determine the characteristics associated with high levels of continuity, and assess the association between continuity and ED visits. DESIGN: Retrospective, observational cohort study of patients at the West Haven VA (WHVA) Primary Care Clinic from March 2011 to February 2012. PATIENTS: The 13,495 patients with established care at the Clinic, having at least one visit, one year before March 2011. MAIN MEASURES: Our exposure variable was continuity of care--a patient seeing their assigned primary care provider (PCP) at each clinic visit. The outcome of interest was having an ED visit. RESULTS: The patients encompassed 42,969 total clinic visits, and 3185 (24%) of them had 15,458 ED visits. In a multivariable logistic regression analysis, patients with continuity of care--at least one visit with their assigned PCP--had lower ED utilization compared to individuals without continuity (adjusted odds ratio [AOR] 0.54; 95% CI: 0.41, 0.71), controlling for frequency of primary care visits, comorbidities, insurance, distance from the ED, and having a trainee PCP assigned. Likewise, the adjusted rate of ED visits was 544/1000 person-year (PY) for patients with continuity vs. 784/1000 PY for patients without continuity (p = 0.001). Compared to patients with low continuity (<33% of visits), individuals with medium (33-50%) and high (>50%) continuity were less likely to utilize the ED. CONCLUSIONS: Strong continuity of care is associated with decreased ED utilization in a PCMH model and improving continuity may help reduce the utilization of non-primary care services.

The need for validation of large administrative databases: Veterans Health Administration ICD-9CM coding of exudative age-related macular degeneration and ranibizumab usage.

We performed a validation study by chart review of data for exudative age-related macular degeneration (eAMD) and, because of the Veterans Administration (VA) therapy policy, ranibizumab usage in the largest electronic medical record system in the USA. We reviewed 5,854 distinct patients who visited an ophthalmology clinic within VA Connecticut from January 2006-December 2008. We randomly selected 98 of 138 distinct eAMD patients and 265 of 5,588 non-eAMD patients who did not receive ranibizumab. International Classification of Diseases, Ninth Revision, Clinical Modification coding of eAMD had an excellent positive predictive value of 97.8% (95% confidence interval (CI), 93.5-99.4%). The national Decision Support System (DSS) had an excellent positive predictive value of 100% (95% CI, 79.9-100%) for ranibizumab. However, the negative predictive value of the DSS dispensed ranibizumab decreased to 67.5 (95% CI, 62.1-72.4) because of a change in the way local values were stored that led to errors. Therefore, validation of clinical information over time in large databases is necessary.

Does Clinical Inertia Vary According to Provider Type?

Despite the well established risks of persistently elevated blood pressure, as well as the benefits of controlling such elevations, hypertension remains underdiagnosed and undertreated. These VA researchers compared the rates of antihypertensive medication intensification between resident physicians, midlevel practitioners, and attending physicians.