A Retrospective Comparison of Creatinine Changes Among Patients Receiving, Not Receiving, and Not Yet Receiving Contrast Administration.

OBJECTIVE: We sought to compare variability in serum creatinine among inpatients in our institution receiving contrast imaging studies and among inpatients not receiving such studies. MATERIALS AND METHODS: This retrospective, single-site, multiple-cohort study in a 550-bed academic medical center in October 2016 used the electronic medical record data to analyze the greatest absolute and relative changes in serum creatinine over periods no longer than 48 hours (1) during the admission for 1,134 patients who did not receive a contrast imaging study, (2) before the earliest contrast study for 155 patients who had not yet had a scheduled contrast examination, and (3) straddling the time when 266 patients received their earliest contrast study. We compared creatinine changes in the first cohort with those in the second and the third using histograms and t tests. RESULTS: Among those who did not receive contrast, 18.3% had a creatinine increase of greater than 0.3 mg/dL, and before contrast, 14.2% had such increases (P = .22). After contrast, 6.4% had increases at least this great (P < .001). Patients with increases in creatinine before contrast tended to have such increases after as well (Pearson's 0.48, P < .001). CONCLUSIONS: Physiological variability may explain the similar increases among patients who did not receive contrast versus patients who had not yet received contrast. Hydration therapy may explain the milder and fewer increases after contrast. Only a randomized clinical trial can determine whether acute kidney injuries are caused by contrast; these results support equipoise for such a trial.

Evaluating Report Text Variation and Informativeness: Natural Language Processing of CT Chest Imaging for Pulmonary Embolism.

OBJECTIVE: The aim of this study was to quantify the variability of language in free text reports of pulmonary embolus (PE) studies and to gauge the informativeness of free text to predict PE diagnosis using machine learning as proxy for human understanding. MATERIALS AND METHODS: All 1,133 consecutive chest CTs with contrast studies performed under a PE protocol and ordered in the emergency department in 2016 were selected from our departmental electronic workflow system. We used commercial text-mining and predictive analytics software to parse and describe all report text and to generate a suite of machine learning rules that sought to predict the "gold standard" radiological diagnosis of PE. RESULTS: There was extensive variation in the length of Findings section and Impression section texts across the reports, only marginally associated with a positive PE diagnosis. A marked concentration of terms was found: for example, 20 words were used in the Findings section of 93% of the reports, and 896 of 2,296 distinct words were each used in only one report's Impression section. In the validation set, machine learning rules had perfect sensitivity but imperfect specificity, a low positive predictive value of 73%, and a misclassification rate of 3%. CONCLUSION: Use of free text reporting was associated with extensive variability in report length and report terms used. Interpretation of the free text was a difficult machine learning task and suggests potential difficulty for human recipients in fully understanding such reports. These results support the prospective assessment of the impact of a fully structured report template with at least some mandatory discrete fields on ease of use of reports and their understanding.

Survival Outcomes Following the Use of Extracorporeal Membrane Oxygenation as a Rescue Technology in Critically Ill Patients: Results From Pennsylvania 2007-2015.

OBJECTIVE: To examine real-world outcomes of survival, length of stay, and discharge destination, among all adult extracorporeal membrane oxygenation admissions in one state over nearly a decade. DESIGN: Retrospective analysis of administrative discharge data. SETTING: State-wide administrative discharge data from Pennsylvania between 2007 and 2015. PATIENTS: All 2,948 consecutive patients billed under a Diagnosis-Related Grouper 3 grouper and in whom a procedural code for extracorporeal membrane oxygenation was present, admitted between the beginning of 2007 and the end of 2015 to hospitals regulated by the state of Pennsylvania. Admitting diagnoses were coded as respiratory, cardiac, cardiac arrest, or uncategorized based on administrative data. MEASUREMENTS AND MAIN RESULTS: Unadjusted in-hospital mortality, length of stay, and discharge destination. Summary statistics and tests of differences by age 65 years or older and by admitting diagnosis were performed. Outcomes by age were plotted using running-mean smoothed graphs. Over the 9-year period, the average observed death rate was 51.7%. Among all survivors, 14.6% went home to self-care and a further 15.2% to home health care. Of all survivors, 43.8% were readmitted within 1 month, and 60.6% within 1 year. Among elderly survivors, readmission rates were 52.3% and 65.5% within 1 month and 1 year, respectively. The likelihood of dying in-hospital increased with age that of being discharged home or to postacute care decreased. CONCLUSIONS: In a "usual clinical practice" setting, short-term outcomes are similar to those observed in clinical trials such as Conventional Ventilation or ECMO for Severe Adult Respiratory Failure, in registries such as extracorporeal life support organization, and in smaller single-site studies. More data on longer term follow-up are needed to allow clinicians to better inform patient selection and care.

Volume-Outcome Relationships in Extracorporeal Membrane Oxygenation: Retrospective Analysis of Administrative Data From Pennsylvania, 2007-2015.

This article seeks to understand whether a volume-outcome relationship exists in adult extracorporeal membrane oxygenation (ECMO). We examined primary administrative discharge data from the Pennsylvanian Health Care Cost Containment Council for all 2,948 consecutive adults treated with ECMO in Pennsylvania between January 1, 2007, and December 31, 2015. We used a well-fitting backwards stepwise logistic regression to obtain patient-level predicted mortality. Number of cases and risk-adjusted mortality was aggregated by calendar quarter and by hospital. Graphical and correlation analysis was used to understand the volume-outcome relationship, focusing separately on the impact of a hospital's current scale (annual volume of adult ECMO) and a hospital's cumulative experience (total cumulative volume of adult ECMO since starting operations). We found that more than 9 years, there was essentially no reduction in overall risk-adjusted mortality in the state. For individual hospitals, once institutions had as few as 50 cases' total experience or performed just 10 cases annually, there was no clear improvement in outcomes. Patients in hospitals with fewer than 50 cases ever performed appeared objectively less sick compared with patients in hospital with more experience. We conclude that there is little evidence of volume-outcome relationships in adult ECMO in this state, although we acknowledge that the absence of clinical chart data limits these conclusions.

Serious Adverse Events Among SPRINT Trial Participants Taking Statins at Baseline.

BACKGROUND: Real-world evidence of statin side effects is potentially biased because statin use is neither randomized nor unblinded. An innovative study design can mitigate these biases. For example, in the recent ASCOT-LLA trial, patient-reported adverse events such as muscle pain and weakness were higher in the non-randomized and non-blinded setting than in the randomized, blinded setting. Less optimally, secondary re-analysis of clinical trials in which statin use is recorded and in which serious adverse events (SAEs) are adjudicated may be conducted. OBJECTIVE: The objective of this study was to evaluate SAEs by statin use at baseline among participants in the SPRINT blood pressure (BP) management trial. METHODS: Unadjusted overall SAE and treatment-related SAE rates by statin use as well as adjusted hazard ratios for statin use were computed in four cohorts [by baseline clinical cardiovascular disease (CVD), by intervention arm]. RESULTS: Statin use at baseline was not associated with higher overall or treatment-related SAE rates among (1) those without pre-existing CVD, regardless of BP arm, nor among (2) those randomized to standard BP management, regardless of pre-existing CVD. Among higher risk patients with existing clinical CVD randomized to intensive BP management, a small but significant increase in overall SAE rate was found among those taking statin at baseline. CONCLUSIONS: In SPRINT, generally statin use was not associated with increased risk of reporting SAEs. Only statin use by higher risk patients was associated with more overall SAEs. Confounding by clinical CVD and the polytherapy of intensive BP management may explain this.

Commercial Online Social Network Data and Statin Side-Effect Surveillance: A Pilot Observational Study of Aggregate Mentions on Facebook.

INTRODUCTION: Surveillance of the safety of prescribed drugs after marketing approval has been secured remains fraught with complications. Formal ascertainment by providers and reporting to adverse-event registries, formal surveys by manufacturers, and mining of electronic medical records are all well-known approaches with varying degrees of difficulty, cost, and success. Novel approaches may be a useful adjunct, especially approaches that mine or sample internet-based methods such as online social networks. METHODS: A novel commercial software-as-a-service data-mining product supplied by Sysomos from Datasift/Facebook was used to mine all mentions on Facebook of statins and stain-related side effects in the US in the 1-month period 9 January 2017 through 8 February 2017. RESULTS: A total of 4.3% of all 25,700 mentions of statins also mentioned typical stain-related side effects. Multiple methodological weaknesses stymie interpretation of this percentage, which is however not inconsistent with estimates that 5-20% of patients taking statins will experience typical side effects at some time. CONCLUSIONS: Future work on pharmacovigilance may be informed by this novel commercial tool, but the inability to mine the full text of a posting poses serious challenges to content categorization.

Evaluation of the "Angelina Jolie Effect" on Screening Mammography Utilization in an Academic Center.

PURPOSE: The aim of this study was to understand the impact on screening mammography at our institution, comparing weekly utilization in the 2 years before and the 2 years after Ms Angelina Jolie disclosed in the New York Times on May 13, 2013, that she had had a prophylactic double mastectomy. METHODS: All 48,110 consecutive screening mammograms conducted at our institution between May 16, 2011, and May 16, 2015, were selected from our electronic medical record system. We used interrupted time series statistical models and graphical methods on utilization data to understand utilization changes before and after Ms Jolie's news. RESULTS: The graphed trend of weekly screening mammogram utilization failed to show changes around the time of interest. Analytical models and statistical tests also failed to show a step change increase or acceleration of utilization around May 2013. However, graphical and time series analyses showed a flattening of utilization in the middle of 2014. CONCLUSIONS: In our well-powered analysis in a large regional breast imaging center, we found no support for the hypothesis that this celebrity news drove increased screening.

Using Social Media, Online Social Networks, and Internet Search as Platforms for Public Health Interventions: A Pilot Study.

OBJECTIVE: To pilot public health interventions at women potentially interested in maternity care via campaigns on social media (Twitter), social networks (Facebook), and online search engines (Google Search). DATA SOURCES/STUDY SETTING: Primary data from Twitter, Facebook, and Google Search on users of these platforms in Los Angeles between March and July 2014. STUDY DESIGN: Observational study measuring the responses of targeted users of Twitter, Facebook, and Google Search exposed to our sponsored messages soliciting them to start an engagement process by clicking through to a study website containing information on maternity care quality information for the Los Angeles market. PRINCIPAL FINDINGS: Campaigns reached a little more than 140,000 consumers each day across the three platforms, with a little more than 400 engagements each day. Facebook and Google search had broader reach, better engagement rates, and lower costs than Twitter. Costs to reach 1,000 targeted users were approximately in the same range as less well-targeted radio and TV advertisements, while initial engagements-a user clicking through an advertisement-cost less than $1 each. CONCLUSIONS: Our results suggest that commercially available online advertising platforms in wide use by other industries could play a role in targeted public health interventions.

Prostate Cancer Care Before and After Medicare Eligibility.

Prior studies suggest Medicare eligibility confers significant and substantial reductions in mortality and beneficial increases in health service utilization. We compared 13,882 patients diagnosed with prostate cancer at ages 63 to 64 years with 14,774 patients diagnosed at ages 65 to 66 (controls) in 2004 to 2007. Compared with controls, patients diagnosed with prostate cancer before Medicare eligibility had no statistically significant or meaningful differences in cancer stage, time to treatment, or type of treatment.

Lung Cancer Care Before and After Medicare Eligibility.

Uninsured and underinsured near-elderly may not have timely investigation, diagnosis, or care of cancer. Prior studies suggest Medicare eligibility confers significant and substantial reductions in mortality and increases in health service utilization. We compared 2245 patients diagnosed with lung cancer at ages 64.5 to 65 years and 2512 patients aged 65 to 65.5 years, with 2492 patients aged 65.5 to 66 years (controls) in 2000 to 2005. Compared with controls, patients diagnosed with lung cancer before Medicare eligibility had no statistically significant differences in cancer stage, time to treatment, type of treatment, and survival. Study power was sufficient to exclude mortality reductions and health service utilization changes of the magnitude found in prior work, suggesting that typically, appropriate lung cancer care may be sought and delivered regardless of insurance status.

Public hospital quality report awareness: evidence from National and Californian Internet searches and social media mentions, 2012.

OBJECTIVES: Publicly available hospital quality reports seek to inform consumers of important healthcare quality and affordability attributes, and may inform consumer decision-making. To understand how much consumers search for such information online on one Internet search engine, whether they mention such information in social media and how positively they view this information. SETTING AND DESIGN: A leading Internet search engine (Google) was the main focus of the study. Google Trends and Google Adwords keyword analyses were performed for national and Californian searches between 1 August 2012 and 31 July 2013 for keywords related to 'top hospital', best hospital', and 'hospital quality', as well as for six specific hospital quality reports. Separately, a proprietary social media monitoring tool was used to investigate blog, forum, social media and traditional media mentions of, and sentiment towards, major public reports of hospital quality in California in 2012. PRIMARY OUTCOME MEASURES: (1) Counts of searches for keywords performed on Google; (2) counts of and (3) sentiment of mentions of public reports on social media. RESULTS: National Google search volume for 75 hospital quality-related terms averaged 610 700 searches per month with strong variation by keyword and by state. A commercial report (Healthgrades) was more commonly searched for nationally on Google than the federal government's Hospital Compare, which otherwise dominated quality-related search terms. Social media references in California to quality reports were generally few, and commercially produced hospital quality reports were more widely mentioned than state (Office of Statewide Healthcare Planning and Development (OSHPD)), or non-profit (CalHospitalCompare) reports. CONCLUSIONS: Consumers are somewhat aware of hospital quality based on Internet search activity and social media disclosures. Public stakeholders may be able to broaden their quality dissemination initiatives by advertising on Google or Twitter and using social media interactively with consumers looking for relevant information.