1. Rare metabolic diseases like TMAU are often underdiagnosed due to their rarity and difficulty diagnosing. However, advances in gastroenterology and metabolic disease research are improving diagnosis and treatment.
2. While TMAU and other rare diseases affect a small percentage of the population, online communities help connect thousands of patients. The definition of "rare" depends on prevalence, and new research may reclassify some conditions.
3. Advances are continually improving our understanding and management of rare conditions. While individual diagnoses remain challenging, increased awareness and research promise better outcomes over time.
NR 512 Effective Communication - tutorialrank.comBartholomew50
For more course tutorials visit
www.tutorialrank.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your practice. What is your primary area where you would use informatics?
Nr 512 Education Organization / snaptutorial.comBaileya136
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your
NR 512 Enhance teaching - snaptutorial.comDavisMurphyA61
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how
Nr 512 Education Organization -- snaptutorial.comDavisMurphyC8
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your practice. What is your primary area where
Nr 512 Education Redefined-snaptutorial.comrobertledwes48
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your practice. What is your primary area where you would use informatics?
What if we never agree on a common health information model?Koray Atalag
In this talk I will touch on some hard problems in health informatics around working with structured data and why we can’t link and reuse them with ease. The essence of the problem is that, while clinicians can perfectly understand each other, IT systems can’t. Traditional IT requires formally defined common terminology, meta-data, data and process definitions. While Medicine is mostly accepted as positive science, yet the great variation in the body of knowledge and practice is often seen as ‘Art’. Ignoring this bit, IT people tend to develop all-inclusive common information models (almost always too complex to implement) and expect everybody adhere to that. Clinicians love to do things a bit differently and of course don’t buy into that! Maybe they are right! Maybe we don’t have to agree on a uniform model at all. This is the basic assumption of the openEHR methodology which I will describe by giving clinical examples. The main premise of this approach is to effectively separate tasks of healthcare and technical professionals. Clinicians can easily define their information needs as they like using visual tools – called Archetypes which are essentially maximal data sets. These computable artefacts, built using a well defined set of technical building blocks, are then fed into the technical environment to integrate data or develop software. Lastly the free web based openEHR Clinical Knowledge Manager portal provides collaborative Archetype development and ensures semantic consistency among different models.
NR 512 Effective Communication - tutorialrank.comBartholomew50
For more course tutorials visit
www.tutorialrank.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your practice. What is your primary area where you would use informatics?
Nr 512 Education Organization / snaptutorial.comBaileya136
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your
NR 512 Enhance teaching - snaptutorial.comDavisMurphyA61
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how
Nr 512 Education Organization -- snaptutorial.comDavisMurphyC8
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your practice. What is your primary area where
Nr 512 Education Redefined-snaptutorial.comrobertledwes48
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Reflect on your own practice. Discuss how informatics is used in your practice. What is your primary area where you would use informatics?
What if we never agree on a common health information model?Koray Atalag
In this talk I will touch on some hard problems in health informatics around working with structured data and why we can’t link and reuse them with ease. The essence of the problem is that, while clinicians can perfectly understand each other, IT systems can’t. Traditional IT requires formally defined common terminology, meta-data, data and process definitions. While Medicine is mostly accepted as positive science, yet the great variation in the body of knowledge and practice is often seen as ‘Art’. Ignoring this bit, IT people tend to develop all-inclusive common information models (almost always too complex to implement) and expect everybody adhere to that. Clinicians love to do things a bit differently and of course don’t buy into that! Maybe they are right! Maybe we don’t have to agree on a uniform model at all. This is the basic assumption of the openEHR methodology which I will describe by giving clinical examples. The main premise of this approach is to effectively separate tasks of healthcare and technical professionals. Clinicians can easily define their information needs as they like using visual tools – called Archetypes which are essentially maximal data sets. These computable artefacts, built using a well defined set of technical building blocks, are then fed into the technical environment to integrate data or develop software. Lastly the free web based openEHR Clinical Knowledge Manager portal provides collaborative Archetype development and ensures semantic consistency among different models.
NR 512 Education Specialist / snaptutorial.comMcdonaldRyan150
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Data/AI driven product development: from video streaming to telehealthXavier Amatriain
Healthcare is different from any other application domain, or is it not? While it is true that there are specific aspects, such as high stakes decisions and a complex regulatory framework, that make healthcare somewhat different, it is also the case that many of the lessons learned from building data-driven products in other domains translate remarcably well into healthcare. This is particularly so because healthcare is also a user facing domain, where users can be both patients or healthcare professionals. Given that data has shown to improve user experience while ensuring quality and scalability, few would argue that healthcare cannot benefit from being much more data-driven than it has traditionally been.
In this talk, I described how this experience building impactful data and AI solutions into user facing products for decades can be leveraged to revolutionize telehealth. At Curai, we combine approaches such as state-of-the-art large language models with expert systems in areas such as NLP, vision, and automated diagnosis to augment and scale doctors, and to improve user experience and healthcare outcomes. We will see some of those applications while analyzing the role of data and ML algorithms in making them possible.
Developing a Virtual Environment for Discharge Planning after Stroke :A Preliminary Study
Interactive Technologies and Games (ITAG) Conference 2014
Health, Disability and Education
Dates: Thursday 16 October 2014 - Friday 17 October 2014
Location: The Council House, NG1 2DT, Nottingham, UK
1. Week 5 Assignment - Case Study Statistical ForecastingDr. TatianaMajor22
1. Week 5 Assignment - Case Study: Statistical Forecasting
Dr. Megan Zobb, a key researcher within the North Luna University Medical Center, has been studying a new variant of a skin disease virus that seems to be surfacing among the North Luna University population. This variant (which has been tentatively named Painful Rash or PR), leads to the formation of surface lesions on an individual's body. These lesions are very similar to small boils or isolated shingles sores. These PR lesions are not necessarily clustered as shingles lesions are, but are isolated across the body.
Insights From Initial Interviews
Megan is initiating some efforts at a preliminary analysis. She has seen 20 initial patients and made several observations about the skin disease. She wants to analyze this initial data before structuring and recommending a more encompassing study.
The signs and symptoms of this disorder usually affect multiple sections of the patient's body. These signs and symptoms may include:
· Pain, burning, numbness or tingling, but pain is always present.
· Sensitivity to touch.
· A red rash that begins a few days after the pain.
· Fluid-filled blisters that break open and crust over.
· Itching.
Some people also experience:
· Fever.
· Headache.
· Sensitivity to light.
· Fatigue.
Pain is always the first symptom of PR. For some, it can be intense. Depending on the location of the pain, it can sometimes be mistaken for a symptom of problems affecting the heart, lungs, or kidneys. Some people experience PR pain without ever developing the rash. The degree of pain that the individual experiences is seemingly proportional to the number of lesions.
Dr. Zobb is extremely concerned that this new variant is especially challenging to the younger population, who are active and like to be outdoors. She has asked you as an analyst and statistician for some assistance in analyzing her initial data. She is not a biostatistician, so she requests that you explain the process you use and your interpretation of the results for each task.
Initial Data Analysis
Dr. Zobb has accumulated some data on an initial set of 20 patients across multiple age groups. She believes that the data suggests younger individuals are affected more than others. She wants you to complete the tasks shown here based on the data below.
For each of the following, provide a detailed explanation of the process you used along with your interpretation of the results. Submit the response in a Word document and attach your Excel spreadsheet to show your calculations (where applicable). Be sure to number each response (e.g., 1.a, 1.b,…).
1. Develop an equation to model the data using a regression analysis approach and explain your calculation process in Excel.
1. Calculate the r-square statistic using Excel. Interpret the meaning of the r-square statistic in this case.
1. Determine three conclusions that address the initial observations and are supported by the regression analysis.
Regression Anal ...
This is a lecture by Joe Lex, MD from the Ghana Emergency Medicine Collaborative. To download the editable version (in PPT), to access additional learning modules, or to learn more about the project, see http://openmi.ch/em-gemc. Unless otherwise noted, this material is made available under the terms of the Creative Commons Attribution Share Alike-3.0 License: http://creativecommons.org/licenses/by-sa/3.0/.
ENDTERMStarted Dec 12 at 604pm Quiz Instructions.docxSALU18
ENDTERM
Started: Dec 12 at 6:04pm
Quiz Instructions
Welcome to Exam #2. As with prior exams, please be in-depth with your answers -- notice the number of points available for each question.
These points are indicative of how involved of a response I require for full credit (this denotes not only amount of writing but quality of that
writing).
Grades will be based on thoughtfulness of responses and tie-in to class discussion/book curriculum. The exam is open book, slides/notes, but not
classmates or other living organisms.
3 pts
HTML Editor
Question 1
Give an example of a product that matches each type of quality and explain why you chose it:
Transcendent:
Product-Based:
Value-Based:
Then discuss which of the types of quality we discussed in class most affect healthcare purchases and why?
Font Sizes Paragraph
p
1 pts
HTML Editor
Question 2
Hoshin planning uses a process called: ___ ___ ___
Do you think Hoshin planning would be effective for hospital management? Why or why not?
Font Sizes Paragraph
p
1 pts
HTML Editor
Question 3
What is the most important role of the marketer in health care? Why?
Font Sizes Paragraph
p
1 pts
HTML Editor
Question 4
Deming says “Eliminate slogans, exhortations, and targets for the work force asking for zero defects and new levels of
productivity.” Do you agree? Why or why not?
Font Sizes Paragraph
p
5 pts
Asks the question, “Are we doing things
right?”(i.e. doing it right the first time)
[ Choose ]
Question 5
Match the portion of CQI to its definition.
Asks the question, “How can we be certain we
do things right the first time, everytime”
[ Choose ]
The outcome or “position” of the 4 types of
improvement
[ Choose ]
Allows us to improve outputs and resource use
[ Choose ]
Asks the question, “Are we doing the right
things?” (i.e. meeting customer expectations)
[ Choose ]
3 pts
HTML Editor
Question 6
Let’s say I want to give you an additional reaction paper to be done during finals week for this class. Using Vroom’s
Expectancy Theory specifically discuss how successful/motivated/etc. you would be in completing this assignment.
Font Sizes Paragraph
p
3 pts
HTML Editor
Question 7
You’re standing on a bridge beside a stranger, watching a large bus careen out of control towards five people. The only way
to stop the bus would be to push the stranger off the bridge thereby changing the direction of the bus. What would you do
and why? Discuss the ethical theory that led to your decision.
...
NR 512 Education Specialist / snaptutorial.comMcdonaldRyan150
For more classes visit
www.snaptutorial.com
NR 512 Week 1 Discussion Integration of Nursing Informatics Skills and Competencies
Integration of Nursing Informatics Skills and Competencies
Data/AI driven product development: from video streaming to telehealthXavier Amatriain
Healthcare is different from any other application domain, or is it not? While it is true that there are specific aspects, such as high stakes decisions and a complex regulatory framework, that make healthcare somewhat different, it is also the case that many of the lessons learned from building data-driven products in other domains translate remarcably well into healthcare. This is particularly so because healthcare is also a user facing domain, where users can be both patients or healthcare professionals. Given that data has shown to improve user experience while ensuring quality and scalability, few would argue that healthcare cannot benefit from being much more data-driven than it has traditionally been.
In this talk, I described how this experience building impactful data and AI solutions into user facing products for decades can be leveraged to revolutionize telehealth. At Curai, we combine approaches such as state-of-the-art large language models with expert systems in areas such as NLP, vision, and automated diagnosis to augment and scale doctors, and to improve user experience and healthcare outcomes. We will see some of those applications while analyzing the role of data and ML algorithms in making them possible.
Developing a Virtual Environment for Discharge Planning after Stroke :A Preliminary Study
Interactive Technologies and Games (ITAG) Conference 2014
Health, Disability and Education
Dates: Thursday 16 October 2014 - Friday 17 October 2014
Location: The Council House, NG1 2DT, Nottingham, UK
1. Week 5 Assignment - Case Study Statistical ForecastingDr. TatianaMajor22
1. Week 5 Assignment - Case Study: Statistical Forecasting
Dr. Megan Zobb, a key researcher within the North Luna University Medical Center, has been studying a new variant of a skin disease virus that seems to be surfacing among the North Luna University population. This variant (which has been tentatively named Painful Rash or PR), leads to the formation of surface lesions on an individual's body. These lesions are very similar to small boils or isolated shingles sores. These PR lesions are not necessarily clustered as shingles lesions are, but are isolated across the body.
Insights From Initial Interviews
Megan is initiating some efforts at a preliminary analysis. She has seen 20 initial patients and made several observations about the skin disease. She wants to analyze this initial data before structuring and recommending a more encompassing study.
The signs and symptoms of this disorder usually affect multiple sections of the patient's body. These signs and symptoms may include:
· Pain, burning, numbness or tingling, but pain is always present.
· Sensitivity to touch.
· A red rash that begins a few days after the pain.
· Fluid-filled blisters that break open and crust over.
· Itching.
Some people also experience:
· Fever.
· Headache.
· Sensitivity to light.
· Fatigue.
Pain is always the first symptom of PR. For some, it can be intense. Depending on the location of the pain, it can sometimes be mistaken for a symptom of problems affecting the heart, lungs, or kidneys. Some people experience PR pain without ever developing the rash. The degree of pain that the individual experiences is seemingly proportional to the number of lesions.
Dr. Zobb is extremely concerned that this new variant is especially challenging to the younger population, who are active and like to be outdoors. She has asked you as an analyst and statistician for some assistance in analyzing her initial data. She is not a biostatistician, so she requests that you explain the process you use and your interpretation of the results for each task.
Initial Data Analysis
Dr. Zobb has accumulated some data on an initial set of 20 patients across multiple age groups. She believes that the data suggests younger individuals are affected more than others. She wants you to complete the tasks shown here based on the data below.
For each of the following, provide a detailed explanation of the process you used along with your interpretation of the results. Submit the response in a Word document and attach your Excel spreadsheet to show your calculations (where applicable). Be sure to number each response (e.g., 1.a, 1.b,…).
1. Develop an equation to model the data using a regression analysis approach and explain your calculation process in Excel.
1. Calculate the r-square statistic using Excel. Interpret the meaning of the r-square statistic in this case.
1. Determine three conclusions that address the initial observations and are supported by the regression analysis.
Regression Anal ...
This is a lecture by Joe Lex, MD from the Ghana Emergency Medicine Collaborative. To download the editable version (in PPT), to access additional learning modules, or to learn more about the project, see http://openmi.ch/em-gemc. Unless otherwise noted, this material is made available under the terms of the Creative Commons Attribution Share Alike-3.0 License: http://creativecommons.org/licenses/by-sa/3.0/.
ENDTERMStarted Dec 12 at 604pm Quiz Instructions.docxSALU18
ENDTERM
Started: Dec 12 at 6:04pm
Quiz Instructions
Welcome to Exam #2. As with prior exams, please be in-depth with your answers -- notice the number of points available for each question.
These points are indicative of how involved of a response I require for full credit (this denotes not only amount of writing but quality of that
writing).
Grades will be based on thoughtfulness of responses and tie-in to class discussion/book curriculum. The exam is open book, slides/notes, but not
classmates or other living organisms.
3 pts
HTML Editor
Question 1
Give an example of a product that matches each type of quality and explain why you chose it:
Transcendent:
Product-Based:
Value-Based:
Then discuss which of the types of quality we discussed in class most affect healthcare purchases and why?
Font Sizes Paragraph
p
1 pts
HTML Editor
Question 2
Hoshin planning uses a process called: ___ ___ ___
Do you think Hoshin planning would be effective for hospital management? Why or why not?
Font Sizes Paragraph
p
1 pts
HTML Editor
Question 3
What is the most important role of the marketer in health care? Why?
Font Sizes Paragraph
p
1 pts
HTML Editor
Question 4
Deming says “Eliminate slogans, exhortations, and targets for the work force asking for zero defects and new levels of
productivity.” Do you agree? Why or why not?
Font Sizes Paragraph
p
5 pts
Asks the question, “Are we doing things
right?”(i.e. doing it right the first time)
[ Choose ]
Question 5
Match the portion of CQI to its definition.
Asks the question, “How can we be certain we
do things right the first time, everytime”
[ Choose ]
The outcome or “position” of the 4 types of
improvement
[ Choose ]
Allows us to improve outputs and resource use
[ Choose ]
Asks the question, “Are we doing the right
things?” (i.e. meeting customer expectations)
[ Choose ]
3 pts
HTML Editor
Question 6
Let’s say I want to give you an additional reaction paper to be done during finals week for this class. Using Vroom’s
Expectancy Theory specifically discuss how successful/motivated/etc. you would be in completing this assignment.
Font Sizes Paragraph
p
3 pts
HTML Editor
Question 7
You’re standing on a bridge beside a stranger, watching a large bus careen out of control towards five people. The only way
to stop the bus would be to push the stranger off the bridge thereby changing the direction of the bus. What would you do
and why? Discuss the ethical theory that led to your decision.
...
Adjusting primitives for graph : SHORT REPORT / NOTESSubhajit Sahu
Graph algorithms, like PageRank Compressed Sparse Row (CSR) is an adjacency-list based graph representation that is
Multiply with different modes (map)
1. Performance of sequential execution based vs OpenMP based vector multiply.
2. Comparing various launch configs for CUDA based vector multiply.
Sum with different storage types (reduce)
1. Performance of vector element sum using float vs bfloat16 as the storage type.
Sum with different modes (reduce)
1. Performance of sequential execution based vs OpenMP based vector element sum.
2. Performance of memcpy vs in-place based CUDA based vector element sum.
3. Comparing various launch configs for CUDA based vector element sum (memcpy).
4. Comparing various launch configs for CUDA based vector element sum (in-place).
Sum with in-place strategies of CUDA mode (reduce)
1. Comparing various launch configs for CUDA based vector element sum (in-place).
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Data and AI
Round table discussion of vector databases, unstructured data, ai, big data, real-time, robots and Milvus.
A lively discussion with NJ Gen AI Meetup Lead, Prasad and Procure.FYI's Co-Found
Adjusting OpenMP PageRank : SHORT REPORT / NOTESSubhajit Sahu
For massive graphs that fit in RAM, but not in GPU memory, it is possible to take
advantage of a shared memory system with multiple CPUs, each with multiple cores, to
accelerate pagerank computation. If the NUMA architecture of the system is properly taken
into account with good vertex partitioning, the speedup can be significant. To take steps in
this direction, experiments are conducted to implement pagerank in OpenMP using two
different approaches, uniform and hybrid. The uniform approach runs all primitives required
for pagerank in OpenMP mode (with multiple threads). On the other hand, the hybrid
approach runs certain primitives in sequential mode (i.e., sumAt, multiply).
Learn SQL from basic queries to Advance queriesmanishkhaire30
Dive into the world of data analysis with our comprehensive guide on mastering SQL! This presentation offers a practical approach to learning SQL, focusing on real-world applications and hands-on practice. Whether you're a beginner or looking to sharpen your skills, this guide provides the tools you need to extract, analyze, and interpret data effectively.
Key Highlights:
Foundations of SQL: Understand the basics of SQL, including data retrieval, filtering, and aggregation.
Advanced Queries: Learn to craft complex queries to uncover deep insights from your data.
Data Trends and Patterns: Discover how to identify and interpret trends and patterns in your datasets.
Practical Examples: Follow step-by-step examples to apply SQL techniques in real-world scenarios.
Actionable Insights: Gain the skills to derive actionable insights that drive informed decision-making.
Join us on this journey to enhance your data analysis capabilities and unlock the full potential of SQL. Perfect for data enthusiasts, analysts, and anyone eager to harness the power of data!
#DataAnalysis #SQL #LearningSQL #DataInsights #DataScience #Analytics
06-04-2024 - NYC Tech Week - Discussion on Vector Databases, Unstructured Data and AI
Discussion on Vector Databases, Unstructured Data and AI
https://www.meetup.com/unstructured-data-meetup-new-york/
This meetup is for people working in unstructured data. Speakers will come present about related topics such as vector databases, LLMs, and managing data at scale. The intended audience of this group includes roles like machine learning engineers, data scientists, data engineers, software engineers, and PMs.This meetup was formerly Milvus Meetup, and is sponsored by Zilliz maintainers of Milvus.
2. About Me
Position:
• Healthcare data scientist
Tools:
• Python (any library that is useful), R,
PySpark, SQL, Julia
• A well-worded email
Hobbies:
• Backpacking
• Long distance runner
• Eating a lot after running
5. Why healthcare data science is so difficult
Privacy
• Privacy is paramount
• Data is very difficult to anonymize
• Privacy is paramount
Data is the new gold
• Healthcare data is extremely valuable
• Models trained on sandboxed data are very valuable and it
is unclear to what extent they can be reverse-engineered
Healthcare Data Is Messy:
• EMR (Electronic Medical Records) data are notoriously
complicated and messy
• Not standardized
• Domain and specialty heavy
6. Problem
Statement
Getting from B -> A
● B: Be able to have a system
such as Google Smart Reply or
Health provider community
forum
● …
● …
● …
● …
● …
● A: Use open-sourced data to
demonstrate intelligent and
contextually health-type
responses
7. Where to get text data for discussions
about your health with medical
professionals that will not violate
anyone’s privacy?
10. The Data
Summary statistics:
• Data gathered from 2014 to 2018
• % of posts by Reddit certified medical professionals: 49.0%
• ~ 30k Threads
• ~ 106k posts (after cleaning)
• ~ 26.3k Users
• ~ 16.3k Medical Professionals
Notes about data:
• More intent in conversations than typically in normal forum or 1-to-1 conversations (especially
Reddit)
• Signs of more male presence, especially in male-specific problems
• Surprisingly tame for Reddit!
14. Scope
Descriptive vs prescriptive:
● Framing the problem for the “patient” as
opposed to trying to solve it for them
Addressing specifics within the
problem:
● Number of speakers
● Change in topic
● Outcome
15. Structuring For
ML
For each thread:
○ Thread start becomes initial query.
Following first-level responses
become become available
responses to initial query
○ E.g.
User1: initial question
User 2: response
User 3: response
User 2: response
User 3: response
User1:
response
16. Looking at Turns
● 96k query/response
pairs
Turns
● How dynamic a
thread/conversation
● Mean: 4.5
● Median: 4.0
● Min: 2
● Max: 106
17. Query:
Alright. Give me a bit. Any
food you can't eat? Egg
products?
Possible Answer 4:
Thanks It's an evolving situation. It seems that it takes awhile for
everyone to get on the same page regarding his medications and
test results. Apparently the positive staph results they got could
have very well been caused by poor procedure. It is all the little
things that are adding up, like the meals, or coming in and asking
a bedridden 93 year old incontinent man, suffering from
dementia, if he needs to get up and go to the rest room and then
leaving when he fails to answer. I'm trying to find a patient
advocate or something now. Supposed to meet with a 'social
worker' as well, but I don't know when.
Possible Answer 1:
I agree with @stephaniecaseys that it is a skin tag. Also not a doc, but I
had mine removed at the dermatologist. It was very quick. It was
messing with my bra strap. They can grow bigger as well, especially if
you have a tendency to mess with it. It will swell.
Possible Answer 3:
You're not the first person to have that happen to. Happens to me just
about every time I'm out in the bush. Unless you are in an area with
known cases of Lyme disease, I doubt it is of any concern.
Possible Answer 2:
I can eat pretty much anything
21. Modeling
Considerations
● Generative modeling is cool, but
has disadvantages:
○ Lots of data
○ May not be precise enough
health context
○ Pretty far away from being able
to augment responses
Keras Blog: Seq2Seq
Cryptic response with generative
approach:
Q: Husband deteriorating before my eyes,
doctors at a loss, no one will help; Reddit
docs, I need you.
A: I don't think this is a single pain is not
a doctor but I have a similar symptoms
and the story
22. Response Retrieval
Approach
● Easier to measure and optimize
than generative approach
● Easier to apply rules by
business user to directed
outcome:
● Allows clinician to automate
follow-up questions that might
often ask in email/phone/visit
exchanges
Changes from Original Paper: The Ubuntu Dialogue Corpus:
A Large Dataset for Research in Unstructured Multi-Turn
Dialogue Systems
Query Response
I’ve had pain in my
knee after running.
Why?
Non-relevant answer 1
Non-relevant answer 2
Relevant answer 1
Non-relevant answer 3
Relevant answer 2
23. Dual Encoder
● Embeddings are initialized with
pre-trained Common Crawl
Glove 840b, 300d and allowed
to train on new utterances &
responses
● 1 hidden layers: one for context,
one for response
● Different choices are made for
output, but often is a binary
probability over up to 10
possible responses
Image credit: The Ubuntu Dialogue Corpus: A Large Dataset for Research in
Unstructured Multi-Turn Dialogue Systems
24. How Good is our Dual Encoder Doing?
Method/Metric Expected
recall due to
chance
TF-IDF LSTM GRU
1 in 2 R@1 50.0% 64.0% 91.1% 92.1%
1 in 5 R@1 20.0% 54.5% 75.5% 79.6%
1 in 10 R@1 10.0% 48.7% 65.1% 68.7%
The above metrics seek to quantify how good the models are at extracting the important
responses. It is a standard metric for response retrieval models.
E.G. “1 in 2 R@1” translates to “At a rate of 1 correct answer out of 2 possible answers,
how what is the % of correct answers if only selecting 1.
Not bad!
25. Future Directions
1. Create intents for all utterances
using a smart indexing from
something like UMLS
2. Refine problem statement to
exploit the potential of data
3. User transfer learning from large
Glove vectors on large corpus of
Reddit comments to get more at
semantics of comments.
4. Experiment with predicting next
response given all thread history
5. Hierarchical attention instead of
RNN as these are showing a lot
of promise.
27. UMLS
Unified Medical Language System
● Brings together many health
vocabularies and standards
● Contains 3 tools:
○ Metathesaurus: Terms and codes from
many vocabularies, including CPT®,
ICD-10-CM, LOINC®, MeSH®, RxNorm,
and SNOMED CT®
○ Semantic Network: Broad categories
(semantic types) and their relationships
(semantic relations)
○ SPECIALIST Lexicon and Lexical
Tools: Natural language processing
tools
28. Modeling
Considerations
● Rule-based:
○ Expensive to build
○ Expensive to maintain
○ Requires precise knowledge
○ E.g.: “If patient has cough,
check for fever”, if in flu and
fever then patient has flu
30. Health Concept Entities
• 61 Total Distinct Concept Entities
• Mean of 20 health entity behaviors per
utterance
31. Identifying Health
Intents
● Ranking of utterances number
of health behavior content
normalized by length
[("Hey, how's your husband doing now? Hope everything is okay.", 0.0),
('I know this is beside your point, but youre 16 and know you have high cholesterol. How? Why?',
1.1831455647530769),
("So why are you posting on here then, if you had two 'real' doctors giving you advice? What answer
are you looking for here? ",
1.2548015599879458),
('How long ago did you change your diet, as in when did you have the kidney stones?',
1.2584445094145018),
('How old is your partner?nnDo you know her diagnosis (ie why they did her surgery)?',
1.2593890745195433),
('How long ago was your thyroid levels checked? Do you have any pain in your abdomen?',
1.2635706685998762),
('Obviously you got blood work done here, any strange findings?',
1.2784053615073325),
("And you want to know if he did the burn on purpose or not? There's absolutely no way of telling,
you'll have to ask him I guess.",
1.311204647849376),
('dysarthria (slurring your words) is not a common side effect of synthroid. Do you take birth control
pills? Do you have a history of migraines? nn Have you had any numbness, weakness, paralysis or
changes in the way you walk?nHave you noticed any other weird things going on with your body
when you have these episodes? nnHave you had trouble with your vision recently? Any trouble
swallowing? Any sudden loss of vision in one eye ever, even a long time ago? ',
1.312565699208395),
('How long have you been suffering? Hope you get rid of dat nasty pain soon.',
1.3201720990394112)]
33. # of Initial Words
vs Quantity of
Responses
● No real relationship between #
of words to number of
responses: that is, using a long
response doesn’t help or hurt
your chances of getting a lot of
input
34. Query:
Because TMAU and many
other metabolic diseases
are very rare. Your other
question is a bit too
general to answer, but
yes, there are advances
all the time in
gastroenterology and
metabolic diseases.
Possible Answer 4:
Could it be tonsil stones perhaps? I have both swollen tonsils
after Is it rare because people don't have the disease, or
because it goes undiagnosed? Or is it difficult to diagnose?
There's a large community over on the MEBO Research site,
with members easily in the thousands. What defines a "rare"
disease? What percentage of the population has to have a
similar or identical diagnosis for it be considered common?
Possible Answer 2:
Thanks a lot! :) You've given me peace of mind.
Possible Answer 3:
Just don't take anymore, you aren't withdrawing from anything. Cut the
b***.
Possible Answer 1:
Curiosity... closure... a few other reasons. I
understand that it would depend on the cause of
death and that if its not a P.E., they might not be
able to tell what it was now depending on how the
body was prepared. But... specifically regarding the
pulmonary embolism... can they at least say
whether it was a P.E. or not?
Editor's Notes
Tried to structure the talk in proportion to the work that has been done somewhat far and also to give some perspective to people outside of healthcare why from my perspective it has unique challenges in data science
Healthcare data is extremely valuable: always consider that before you share your data with just any app.
You would think that Kaiser Permanente would be above using Reddit data. Not So!
You are highly susceptible I think to Apophenia: seeing patterns in random data
Two points to make:
1) The consideration of know who is who (medical professional or not)
2) Making sure the optimization is happening more on the clinician side as opposed to the ”customer” or
Becomes important as you consider how long the thread goes on
Trimethylaminuria is a disorder in which the body is unable to break down trimethylamine, a chemical compound that has a pungent odor. Trimethylamine has been described as smelling like rotting fish, rotting eggs, garbage, or urine
A Dual Encoder Sequence to Sequence Model for Open-Domain Dialogue Modeling: https://arxiv.org/pdf/1710.10520.pdf
E.g. “1 in 2 R@1” means for 1 correct recall of the recommendation out of 2 total recommendation where 1 in 2 are correct
Expensive to build:
Some things are easy: “If patient has cough, check for fever”, if in flu and fever then patient has flue
Health-care may be one of the worst examples for this however, since any one manifestation of a symptom probably means a bunch of things. In fact
Trimethylaminuria is a disorder in which the body is unable to break down trimethylamine, a chemical compound that has a pungent odor. Trimethylamine has been described as smelling like rotting fish, rotting eggs, garbage, or urine