Mining social media data to assess the risk of skin and soft tissue infections from allergen immunotherapy.

BACKGROUND: Allergen immunotherapy (AIT) treatment for allergic rhinitis and asthma is used by 2.6 million Americans annually. Clinical and sterility testing studies identify no risk of contamination or infection from extracts prepared using recommended aseptic techniques, but regulatory concerns persist. Social media can be used to investigate rare adverse effects not captured by traditional studies. OBJECTIVE: We sought to investigate large social media databases for suggestion of AIT skin and soft tissue infection (SSTI) risk and compare this risk to a comparator procedure with a sterile pharmaceutical. METHODS: We analyzed US-restricted data from more than 10 common text-based social media platforms including Facebook, Twitter, and Reddit between 2012 and 2016. We used natural language processing (NLP) to identify posts related to AIT and, separately, influenza vaccination (comparator procedure). NLP was followed by manual review to identify posts suggesting a possible SSTI associated with either AIT or influenza vaccination. SSTI frequencies with 95% CIs were compared. RESULTS: We identified 25,126 AIT posts, which were matched by social media platform to 25,126 influenza vaccination-related posts. NLP identified 4088 (16.3%) AIT posts that required manual review, with 6 posts (0.02%; 95% CI, 0.005%-0.043%) indicative of possible AIT-related SSTI. NLP identified 2689 (10.7%) influenza posts that required manual review, with 7 posts (0.03%; 95% CI, 0.007%-0.048%) indicative of possible influenza vaccination-related SSTI. CONCLUSIONS: Social media data suggest that SSTI from AIT and influenza vaccination are equally rare events. Given that AIT's SSTI risk appears comparable to the risk using a sterile pharmaceutical based on social media data, current aseptic technique procedures seem safe.

Mining fall-related information in clinical notes: Comparison of rule-based and novel word embedding-based machine learning approaches.

BACKGROUND: Natural language processing (NLP) of health-related data is still an expertise demanding, and resource expensive process. We created a novel, open source rapid clinical text mining system called NimbleMiner. NimbleMiner combines several machine learning techniques (word embedding models and positive only labels learning) to facilitate the process in which a human rapidly performs text mining of clinical narratives, while being aided by the machine learning components. OBJECTIVE: This manuscript describes the general system architecture and user Interface and presents results of a case study aimed at classifying fall-related information (including fall history, fall prevention interventions, and fall risk) in homecare visit notes. METHODS: We extracted a corpus of homecare visit notes (n = 1,149,586) for 89,459 patients from a large US-based homecare agency. We used a gold standard testing dataset of 750 notes annotated by two human reviewers to compare the NimbleMiner's ability to classify documents regarding whether they contain fall-related information with a previously developed rule-based NLP system. RESULTS: NimbleMiner outperformed the rule-based system in almost all domains. The overall F- score was 85.8% compared to 81% by the rule based-system with the best performance for identifying general fall history (F = 89% vs. F = 85.1% rule-based), followed by fall risk (F = 87% vs. F = 78.7% rule-based), fall prevention interventions (F = 88.1% vs. F = 78.2% rule-based) and fall within 2 days of the note date (F = 83.1% vs. F = 80.6% rule-based). The rule-based system achieved slightly better performance for fall within 2 weeks of the note date (F = 81.9% vs. F = 84% rule-based). DISCUSSION & CONCLUSIONS: NimbleMiner outperformed other systems aimed at fall information classification, including our previously developed rule-based approach. These promising results indicate that clinical text mining can be implemented without the need for large labeled datasets necessary for other types of machine learning. This is critical for domains with little NLP developments, like nursing or allied health professions.

NimbleMiner: An Open-Source Nursing-Sensitive Natural Language Processing System Based on Word Embedding.

This study develops and evaluates an open-source software (called NimbleMiner) that allows clinicians to interact with word embedding models with a goal of creating lexicons of similar terms. As a case study, the system was used to identify similar terms for patient fall history from homecare visit notes (N = 1 149 586) extracted from a large US homecare agency. Several experiments with parameters of word embedding models were conducted to identify the most time-effective and high-quality model. Models with larger word window width sizes (n = 10) that present users with about 50 top potentially similar terms for each (true) term validated by the user were most effective. NimbleMiner can assist in building a thorough vocabulary of fall history terms in about 2 hours. For domains like nursing, this approach could offer a valuable tool for rapid lexicon enrichment and discovery.

Improving patient prioritization during hospital-homecare transition: A pilot study of a clinical decision support tool.

Patient admission to homecare is a complex process. Medicare policy requires that all patients receive a first home visit within 48 hr after the referral is received at the homecare agency. For unstable or high risk patients, waiting 48 hr to be seen by homecare nurses may not be safe. In this pilot study we tested an innovative clinical decision support tool (called PREVENT), designed to identify patients who may need to be prioritized for early homecare visits. The study was conducted in 2016 at a large homecare agency in the Northeastern US with 176 patients admitted to homecare from the hospital. In the control phase (n = 90 patients), we calculated the PREVENT priority score (indicative of high or medium/low first nursing visit priority) but did not share the score with the homecare intake nurses who influence visit scheduling. In the experimental phase, the PREVENT score was shared with the homecare intake nurses (n = 86 patients). During the experimental phase, high-risk patients received their first homecare nursing visit about one-half a day sooner than in the control phase (1.8 days vs. 2.2 days). Rehospitalizations from homecare decreased by 9.4% between the control (21.1%) and experimental phases (11.7%). This pilot study of patient prioritization showed promising results: high priority patients received their first homecare visit sooner and overall rehospitalization rates were lower.

Patient-centered care via health information technology: a qualitative study with experts from Israel and the U.S.

Although patient-centered care (PCC) is one of the cornerstones of modern healthcare, the role that health information technology (HIT) plays in supporting PCC remains unclear. In this qualitative study, we interviewed academic and clinical experts from the US and Israel to understand to what extent current HIT systems are supportive of PCC and how PCC should be supported by HIT in the future. A maximum variation sampling approach was used to identify nine experts in both HIT and PCC from clinical and academic settings in Israel and the US. A qualitative descriptive method was used to analyze the interviews and identify major themes. Experts suggested that patient ownership of their disease is a core component of PCC. The majority of the experts agreed that in both Israel and the US, the current situation of PCC implementation is relatively poor. However, HIT should play an important role in making patients owners of their health and treatment and helping providers in delivering better PCC. Central domains of PCC via HIT were providing clear information and support for patients and promoting care that is based on patient values and preferences.

Extracting Alcohol and Substance Abuse Status from Clinical Notes: The Added Value of Nursing Data.

We applied an open source natural language processing (NLP) system "NimbleMiner" to identify clinical notes with mentions of alcohol and substance abuse. NimbleMiner allows users to rapidly discover clinical vocabularies (using word embedding model) and then implement machine learning for text classification. We used a large inpatient dataset with over 50,000 intensive care unit admissions (MIMIC II). Clinical notes included physician-written discharge summaries (n = 51,201) and nursing notes (n = 412,343). We first used physician-written discharge summaries to train the system's algorithm and then added nursing notes to the physician-written discharge summaries and evaluated algorithms prediction accuracy. Adding nursing notes to the physician-written discharge summaries resulted in almost two-fold vocabulary expansion. NimbleMiner slightly outperformed other state-of-the-art NLP systems (average F-score = .84), while requiring significantly less time for the algorithms development.: Our findings underline the importance of nursing data for the analysis of electronic patient records.

Nonpainful wide-area compression inhibits experimental pain.

Compression therapy, a well-recognized treatment for lymphoedema and venous disorders, pressurizes limbs and generates massive non-noxious afferent sensory barrages. The aim of this study was to study whether such afferent activity has an analgesic effect when applied on the lower limbs, hypothesizing that larger compression areas will induce stronger analgesic effects, and whether this effect correlates with conditioned pain modulation (CPM). Thirty young healthy subjects received painful heat and pressure stimuli (47°C for 30 seconds, forearm; 300 kPa for 15 seconds, wrist) before and during 3 compression protocols of either SMALL (up to ankles), MEDIUM (up to knees), or LARGE (up to hips) compression areas. Conditioned pain modulation (heat pain conditioned by noxious cold water) was tested before and after each compression protocol. The LARGE protocol induced more analgesia for heat than the SMALL protocol (P < 0.001). The analgesic effect interacted with gender (P = 0.015). The LARGE protocol was more efficient for females, whereas the MEDIUM protocol was more efficient for males. Pressure pain was reduced by all protocols (P < 0.001) with no differences between protocols and no gender effect. Conditioned pain modulation was more efficient than the compression-induced analgesia. For the LARGE protocol, precompression CPM efficiency positively correlated with compression-induced analgesia. Large body area compression exerts an area-dependent analgesic effect on experimental pain stimuli. The observed correlation with pain inhibition in response to robust non-noxious sensory stimulation may suggest that compression therapy shares similar mechanisms with inhibitory pain modulation assessed through CPM.