MedJEx: A Medical Jargon Extraction Model with Wiki's Hyperlink Span and Contextualized Masked Language Model Score.

This paper proposes a new natural language processing (NLP) application for identifying medical jargon terms potentially difficult for patients to comprehend from electronic health record (EHR) notes. We first present a novel and publicly available dataset with expert-annotated medical jargon terms from 18K+ EHR note sentences (MedJ). Then, we introduce a novel medical jargon extraction (MedJEx) model which has been shown to outperform existing state-of-the-art NLP models. First, MedJEx improved the overall performance when it was trained on an auxiliary Wikipedia hyperlink span dataset, where hyperlink spans provide additional Wikipedia articles to explain the spans (or terms), and then fine-tuned on the annotated MedJ data. Secondly, we found that a contextualized masked language model score was beneficial for detecting domain-specific unfamiliar jargon terms. Moreover, our results show that training on the auxiliary Wikipedia hyperlink span datasets improved six out of eight biomedical named entity recognition benchmark datasets. Both MedJ and MedJEx are publicly available.

Inadequate diversity of information resources searched in US-affiliated systematic reviews and meta-analyses: 2005-2016.

OBJECTIVE: Systematic reviews and meta-analyses (SRMAs) rely upon comprehensive searches into diverse resources that catalog primary studies. However, since what constitutes a comprehensive search is unclear, we examined trends in databases searched from 2005-2016, surrounding the publication of search guidelines in 2013, and associations between resources searched and evidence of publication bias in SRMAs involving human subjects. STUDY DESIGN: To ensure comparability of included SRMAs over the 12 years in the face of a near 100-fold increase of international SRMAs (mainly genetic studies from China) during this period, we focused on USA-affiliated SRMAs, manually reviewing 100 randomly selected SRMAs from those published in each year. After excluding articles (mainly for inadequate detail or out-of-scope methods), we identified factors associated with the databases searched, used network analysis to see which resources were simultaneously searched, and used logistic regression to link information sources searched with a lower chance of finding publication bias. RESULTS: Among 817 SRMA articles studied, the common resources used were Medline (95%), EMBASE (44%), and Cochrane (41%). Methods journal SRMAs were most likely to use registries and grey literature resources. We found substantial co-searching of resources with only published materials, and not complemented by searches of registries and the grey literature. The 2013 guideline did not substantially increase searching of registries and grey literature resources to retrieve primary studies for the SRMAs. When used to augment Medline, Scopus (in all SRMAs) and ClinicalTrials.gov (in SRMAs with safety outcomes) were negatively associated with publication bias. CONCLUSIONS: Even SRMAs that search multiple sources tend to search similar resources. Our study supports searching Scopus and CTG in addition to Medline to reduce the chance of publication bias.

Clinical and biological moderators of response to naltrexone in alcohol dependence: a systematic review of the evidence.

AIM: The goal of this systematic review was to identify moderators of naltrexone efficacy in the treatment of alcohol dependence. METHODS: We searched Pubmed, CINHAL, Embase, PsycINFO and the Cochrane Library from 1990 to April 2012 and reference lists of pertinent review articles, which yielded 622 trial, pooled analysis and review articles. Using pre-established eligibility criteria, two reviewers independently determined whether abstracts contained evidence of demographic or biological characteristics, i.e. moderators, influencing naltrexone response in alcohol dependence. We assessed each publication for risk of bias and evaluated the strength of the body of evidence for each moderator. RESULTS: Twenty-eight publications (on 20 studies) met criteria for data synthesis. These included 26 publications from 12 randomized, placebo-controlled trials, three non-randomized, non-placebo studies and one randomized, non-placebo study. In addition, there were two publications from pooled analyses of four randomized, placebo-controlled trials. Family history of alcohol problems and the Asn40Asp polymorphism of the µ-opioid receptor gene showed a positive association with efficacy in four of five and three of five studies, respectively. Other moderators reported to be associated with efficacy included male sex (two of five studies), pre-treatment drinking (two of two studies) and high craving (two of five studies). However, the overall risk of bias in the published literature is high. CONCLUSIONS: The identification of naltrexone-responsive alcohol-dependent patients is still in development. Studies to date point to two potential moderators-family history and presence of the OPRM1 Asn40Asp polymorphism-as having the strongest evidence. However, the data to date is still insufficient to recommend that any moderator be used in determining clinical treatment.

Modifying ICD-9-CM coding of secondary diagnoses to improve risk-adjustment of inpatient mortality rates.

OBJECTIVE: To assess the effect on risk-adjustment of inpatient mortality rates of progressively enhancing administrative claims data with clinical data that are increasingly expensive to obtain. Data Sources. Claims and abstracted clinical data on patients hospitalized for 5 medical conditions and 3 surgical procedures at 188 Pennsylvania hospitals from July 2000 through June 2003. METHODS: Risk-adjustment models for inpatient mortality were derived using claims data with secondary diagnoses limited to conditions unlikely to be hospital-acquired complications. Models were enhanced with one or more of 1) secondary diagnoses inferred from clinical data to have been present-on-admission (POA), 2) secondary diagnoses not coded on claims but documented in medical records as POA, 3) numerical laboratory results from the first hospital day, and 4) all available clinical data from the first hospital day. Alternative models were compared using c-statistics, the magnitude of errors in prediction for individual cases, and the percentage of hospitals with aggregate errors in prediction exceeding specified thresholds. RESULTS: More complete coding of a few under-reported secondary diagnoses and adding numerical laboratory results to claims data substantially improved predictions of inpatient mortality. Little improvement resulted from increasing the maximum number of available secondary diagnoses or adding additional clinical data. CONCLUSIONS: Increasing the completeness and consistency of reporting a few secondary diagnosis codes for findings POA and merging claims data with numerical laboratory values improved risk adjustment of inpatient mortality rates. Expensive abstraction of additional clinical information from medical records resulted in little further improvement.

Piloting nursing-sensitive hospital care measures in Massachusetts.

Under the umbrella of the Massachusetts Hospital Association and Massachusetts Organization of Nurse Executives Patients First Initiative, Massachusetts hospitals tested a subset of NQF-endorsed nursing-sensitive care measures in 2006. In this report, we describe the pilot test, report on pilot test measure data, summarize participant feedback on the tested measures, and offer observations on lessons learned from the pilot test.

Combining administrative and clinical data to stratify surgical risk.

OBJECTIVE: To evaluate whether administrative claims data (ADM) from hospital discharges can be transformed by present-on-admission (POA) codes and readily available clinical data into a refined database that can support valid risk stratification (RS) of surgical outcomes. SUMMARY BACKGROUND DATA: ADM from hospital discharges have been used for RS of medical and surgical outcomes, but results generally have been viewed with skepticism because of limited clinical information and questionable predictive accuracy. METHODS: We used logistic regression analysis to choose predictor variables for RS of mortality in abdominal aortic aneurysm repair, coronary artery bypass graft surgery, and craniotomy, and for RS of 4 postoperative complications (ie, physiologic/metabolic derangement, respiratory failure, pulmonary embolism/deep vein thrombosis, and sepsis) after selected operations. RS models were developed for age only (Age model), ADM only (ADM model), ADM enhanced with POA codes for secondary diagnoses (POA-ADM model), POA-ADM supplemented with admission laboratory data (Laboratory model), Laboratory model supplemented with admission vital signs and additional laboratory data (VS model), VS model supplemented with key clinical findings abstracted from medical records (KCF model), and KCF model supplemented with composite clinical scores (Full model). Models were evaluated using c-statistics, case-based errors in predictions, and measures of hospital-based systematic bias. RESULTS: The addition of POA codes and numerical laboratory results to ADM was associated with substantial improvements in all measures of analytic performance. In contrast, the addition of difficult-to-obtain key clinical findings resulted in only small improvements in predictions. CONCLUSIONS: Enhancement of ADM with POA codes and readily available laboratory data can efficiently support accurate risk-stratified measurements of clinical outcomes in surgical patients.

Enhancement of claims data to improve risk adjustment of hospital mortality.

CONTEXT: Comparisons of risk-adjusted hospital performance often are important components of public reports, pay-for-performance programs, and quality improvement initiatives. Risk-adjustment equations used in these analyses must contain sufficient clinical detail to ensure accurate measurements of hospital quality. OBJECTIVE: To assess the effect on risk-adjusted hospital mortality rates of adding present on admission codes and numerical laboratory data to administrative claims data. DESIGN, SETTING, AND PATIENTS: Comparison of risk-adjustment equations for inpatient mortality from July 2000 through June 2003 derived by sequentially adding increasingly difficult-to-obtain clinical data to an administrative database of 188 Pennsylvania hospitals. Patients were hospitalized for acute myocardial infarction, congestive heart failure, cerebrovascular accident, gastrointestinal tract hemorrhage, or pneumonia or underwent an abdominal aortic aneurysm repair, coronary artery bypass graft surgery, or craniotomy. MAIN OUTCOME MEASURES: C statistics as a measure of the discriminatory power of alternative risk-adjustment models (administrative, present on admission, laboratory, and clinical for each of the 5 conditions and 3 procedures). RESULTS: The mean (SD) c statistic for the administrative model was 0.79 (0.02). Adding present on admission codes and numerical laboratory data collected at the time of admission resulted in substantially improved risk-adjustment equations (mean [SD] c statistic of 0.84 [0.01] and 0.86 [0.01], respectively). Modest additional improvements were obtained by adding more complex and expensive to collect clinical data such as vital signs, blood culture results, key clinical findings, and composite scores abstracted from patients' medical records (mean [SD] c statistic of 0.88 [0.01]). CONCLUSIONS: This study supports the value of adding present on admission codes and numerical laboratory values to administrative databases. Secondary abstraction of difficult-to-obtain key clinical findings adds little to the predictive power of risk-adjustment equations.

The hazards of using administrative data to measure surgical quality.

Administrative claims data have been used to measure risk-adjusted clinical outcomes of hospitalized patients. These data have been criticized because they cannot differentiate risk factors present at the time of admission from complications that occur during hospitalization. This paper illustrates how valid risk-adjustment can be achieved by enhancing administrative data with a present-on-admission code, admission laboratory data, and admission vital signs. Examples are presented for inpatient mortality rates following craniotomy and rates of postoperative sepsis after elective surgical procedures. Administrative claims data alone yielded a risk-adjustment model with 10 variables and a C-statistic of 0.891 for mortality after craniotomy, and a model with 18 variables and a C-statistic of 0.827 for postoperative sepsis. In contrast, the combination of administrative data and clinical data abstracted from medical records increased the number of variables in the craniotomy model to 21 with a C-statistic of 0.923, and the number of variables in the postoperative sepsis model to 29 with a C-statistic of 0.858. Use of only administrative data resulted in unacceptable amounts of systematic bias in 24 per cent of hospitals for craniotomy and 19 per cent of hospitals for postoperative sepsis. Addition of a present-on-admission code, laboratory data, and vital signs reduced the percentage of hospitals with unacceptable bias to two percent both for craniotomy and for postoperative sepsis. These illustrations demonstrate suboptimal risk stratification with administrative claims data only, but show that present-on-admission coding combined with readily available laboratory data and vital signs can support accurate risk-adjustment for the assessment of surgical outcomes.

n-3 Fatty acids from fish or fish-oil supplements, but not alpha-linolenic acid, benefit cardiovascular disease outcomes in primary- and secondary-prevention studies: a systematic review.

Studies on the relation between dietary n-3 fatty acids (FAs) and cardiovascular disease vary in quality, and the results are inconsistent. A systematic review of the literature on the effects of n-3 FAs (consumed as fish or fish oils rich in eicosapentaenoic acid and docosahexaenoic acid or as alpha-linolenic acid) on cardiovascular disease outcomes and adverse events was conducted. Studies from MEDLINE and other sources that were of > or =1 y in duration and that reported estimates of fish or n-3 FA intakes and cardiovascular disease outcomes were included. Secondary prevention was addressed in 14 randomized controlled trials (RCTs) of fish-oil supplements or of diets high in n-3 FAs and in 1 prospective cohort study. Most trials reported that fish oil significantly reduced all-cause mortality, myocardial infarction, cardiac and sudden death, or stroke. Primary prevention of cardiovascular disease was reported in 1 RCT, in 25 prospective cohort studies, and in 7 case-control studies. No significant effect on overall deaths was reported in 3 RCTs that evaluated the effects of fish oil in patients with implantable cardioverter defibrillators. Most cohort studies reported that fish consumption was associated with lower rates of all-cause mortality and adverse cardiac outcomes. The effects on stroke were inconsistent. Evidence suggests that increased consumption of n-3 FAs from fish or fish-oil supplements, but not of alpha-linolenic acid, reduces the rates of all-cause mortality, cardiac and sudden death, and possibly stroke. The evidence for the benefits of fish oil is stronger in secondary- than in primary-prevention settings. Adverse effects appear to be minor.

Effects of statins on vascular structure and function: a systematic review.

PURPOSE: Statins reduce cardiovascular events by more than can be explained by their effects on lipids. We conducted a systematic review of how statins affect vascular structure and function, differences among statins, and correlations between the effects of statins on vascular outcomes and either lipid levels or cardiovascular outcomes. METHODS: We primarily searched MEDLINE (1980 to March 2004) to identify all studies with at least 10 subjects that reported the effects of currently available statins on coronary artery stenosis, carotid intima-media thickness, and endothelial function (excluding studies of drug combinations and subjects with organ transplants). Meta-analyses were performed when feasible. RESULTS: Statins decrease the progression and increase the regression of coronary artery lesions and luminal narrowing. Compared with placebo, statins decrease the likelihood of coronary artery restenosis (summary risk ratio = 0.85; 95% confidence interval: 0.77 to 0.95). Statins appear to slow the progression of carotid artery intima-media thickness. Although the effect of statins on coronary endothelial function is uncertain, statins appear to improve peripheral endothelial function. There is no conclusive evidence to suggest that individual statins differ in their effects on these outcomes. Studies generally found weak or no correlation between the effects of statins on vascular outcomes and lipid levels. No study showed a correlation between vascular effect and clinical outcome. CONCLUSION: Statins slow the progression of, and may reverse, atherosclerosis. The magnitude of these effects, however, is small compared with the effects of statins on cardiovascular events. Statins also improve measures of vascular function, which may contribute to their clinical benefits. There is insufficient evidence to suggest that individual statins differ in their vascular effects.

Effects of statins on nonlipid serum markers associated with cardiovascular disease: a systematic review.

BACKGROUND: Statins reduce cardiovascular events to a greater extent than can be explained by their effect on lipids. Several studies have attempted to elucidate mechanisms by which statins reduce cardiovascular risk. PURPOSE: To summarize the effects of statins on nonlipid serum markers and to correlate statins' effect on serum markers with lipid levels and cardiovascular outcomes. DATA SOURCES: MEDLINE (1980 to 2003) search limited to English-language articles. STUDY SELECTION: Studies reporting original data in at least 10 participants on the effect of statins on outcomes of interest, excluding studies of cerivastatin, drug combinations, and patients with organ transplants. DATA EXTRACTION: Study design, sample size, treatment, and outcome data extracted on the basis of preestablished criteria. When appropriate, meta-analysis was performed by using a random-effects model. DATA SYNTHESIS: All statins are effective at lowering C-reactive protein levels, and the effect is not dose-dependent. Studies do not demonstrate a correlation between statins' effects on C-reactive protein levels and on lipids or cardiovascular outcomes. Statins do not affect fibrinogen levels, and limited data suggest little effect on lipid oxidation, tissue plasminogen activator, or plasminogen activator inhibitor. Platelet aggregation data are inconclusive. CONCLUSIONS: Among nonlipid serum markers examined, only C-reactive protein levels are statistically significantly affected by statins. These findings suggest that statin-mediated anti-inflammatory effects may contribute to the ability of statins to reduce risk for cardiovascular disease. Overall, however, available data are insufficient to support recommendations for using nonlipid serum markers in decisions regarding statin therapy for individual patients.

Linking pharmacoeconomic analyses to results of systematic review and meta-analysis.

Pharmacoeconomic analysis applies quantitative modeling to the assessment of the clinical and economic impact of new drugs. Users of pharmacoeconomic analysis include government agencies, government payers and policy makers, private payers (including managed care organizations) and pharmaceutical companies. Pharmacoeconomic analyses can aid policy decisions, provide support for better allocation of scarce resources and assist clinical decisions. Since pharmacoeconomic analyses can have a wide impact, it is important that they are based upon reliable data. Well-conducted systematic reviews and meta-analyses can provide high quality data to pharmacoeconomic analyses, with considerable synergy achieved by combining these two powerful methodologies. An overview of systematic review and meta-analysis are presented and some examples of their use in pharmacoeconomics are described.