tobiyusuf

[tobiyusuf]

The Importance of Information Visualization in Healthcare

Throughout my education, I have had to give and listen to presentations. Many of them did not convey data affectively. The presentation either had too much words on a slide while the presenter read off what was on the slides. Others had cluttered images and quotes. Giving presentations in this format not only makes the presentation not interesting, it prevents the main message from getting across. This is why visualizing data affectively is important, it gets the main across and calls for action based on the data.

Tips that were given to me when it comes to presenting ranged from cutting down words on a slide, only present relative data, and to grab the audiences attention through storytelling and bolding main ideas on a slide. These tips have been proven to be successful in presenting data in all fields of study. Unfortunately, there is lack of applying the principles of information visualization in many areas such as Healthcare. Due to the the rise of electronic health records (EHR) and big data, It is difficult to get a quick assessment of a patients state of health and data is organized poorly in some cases. It is not readily easy to know a patient’s outcome from looking at data. Below is an example of a written health record and an EHR:

The information is not displayed here does not easily grab the attention of the doctor reading it and is harder to formulate a treatment plan. To address these issues, Zhang et al. provides tips on how to make data visualization in EHR more affective.

Zhang et al. use the Five Ws (who, when, what, where, and why)  which are necessary to provide a compelling story from data. The who and what provides the doctor with information about the patient’s history which include symptoms and their severity, diagnosis, past procedures and treatments. Where refers to where the who and what information are located on the body. The when and why information explain causal relationships between the symptoms and diagnosis. This data help decide what treatments should done and cast diagnoses if it is not known. Below is a image that visualizes the 5 whys:

(a) demonstrates a hierarchical radial display and a body outline. The body demonstrates the where and who information, the different nodes show the what information, the colors demonstrate severity and draw the doctor to focus more on there areas. (b) demonstrates the when and why information by showing a sequential display of events that the patient has gone through. The doctor can pay attention more to the more severe areas and formulate a patient plan more easily. This image is more affective in displaying patient data. The image below shows an event more integrated display of patient data throughout the course of 4 visits:

This data present collaboration between four different doctors in different fields. The work flow diagram demonstrate when and why information and the hierarchical diagram shows how the patients symptoms and severity has progressed throughout the visits. The visualization of the patients data is more organized and easier to read. Zhang et al. presented this data to a few doctors and most of them found it easier to find the most severe diseases of the patient, the locations of the diseases, saw the links in the workflow digram useful, and the ratings of the diagrams from a scale of 0 to 10 ranged from 5 to 10. Overall, the informational visualization of the patients data was more informative and compelling than the traditional EHR.

Even though this data is shown in an affective way, the data will be presented better with a story. Showing the data alone would not be as effective if it is not backed up with a story. 

If I were to work in the healthcare setting, I would want to have patient data presented in this way because it will be easier for me to analyze the data an take action on it. In addition, I think that the patient should also receive information about themselves through visuals like these. When I go to doctor visits, I mostly receive information about myself verbally and not so visually. I would like to get more visuals on what my symptoms are. I hope to improve how patient data is presented in the future.

Sources:

https://www.sas.com/en_us/insights/big-data/data-visualization.html#dmimportance

http://ieeexplore.ieee.org/document/6523038/#full-text-section

[tobiyusuf]

Process Analytics in Healthcare

Process analytics is a process of creating a plan on executing a certain project. It is aimed to improve processes that occur in a business. Businesses use process analytics when it come to planning and redesigning how to promote products, detect fraud, appease customers, improve production of products, etc. To design a plan, one must know where the process starts and ends, sub processes that occur within the process, and everyone who participates in the process. Several models are used to help design and visualize processes such as process maps and workflow models. A type of process that I will talk about is the transition of care in Emergency Healthcare.

Processes that occur in Emergency Medical Services are very complex. Processes involve many participants such as the patient, family members of the patient, Emergency Medical Technicians (EMTs), nurses, doctors, hospital staff, and much more. The process for providing care for a patient may start from when 911 is called for the patient and end when the patient is fully cared for (or not so). Sub-processes that occur when caring for the patient include the EMTs process of caring for the patient, the hospital staff caring for the patient in the emergency room, and the patient’s recovery in the hospital. What sub-processes that occur depend on what the patient’s illness and/or mechanisms of injuries are. This is why healthcare processes can be very complex.

Furthermore, lots of research are being done to improve the health care process. Since the rise of electronic healthcare records, more big data is being collected. Process analytics help to analyze this data and to make improvements. In A guide for the application of analytics on healthcare processes: A dynamic view on patient pathways, Lismont et al. attempt to create workflow diagrams on the process of patients receiving care for diabetes. They used data mining technique to pick out the similarities in the steps that patients take in their course of care. They generated one that only represent the workflow of 20% of the patients and a map of drugs that only 50% of the patients follow. This shows the complexity of this process. Patients usually have their unique plan of care and not everyone follows the same steps. Lismont et al. suggest that more studies are done to improve abstraction techniques while looking through data, include how much each event will cost, find ways to better predict patient outcome and medical services provided.

When creating a plan for a process, one must be aware about how complex a process can be. a process will most likely contain sub-process and not every business or patient will follow the same steps of a process as another patient or business. In addition, creating workflow diagrams and process map are important. It proved a better way to visualize a process and in some areas, it is required. Further analysis can be done in order to make improvements in the process and reduce cost. I would like to do more studies on how to improve patient care, reduce cost in each event needed for the patient, and to better organize the processes that take place health care.

Sources:

http://data-informed.com/8-business-process-analytics-every-manager-should-know/

https://ac.els-cdn.com/S0010482516301986/1-s2.0-S0010482516301986-main.pdf?_tid=c969bda8-cc7e-11e7-9a53-00000aacb360&acdnat=1511023237_b3baf2c5bd0ec5759f523431bcbd4c11

[tobiyusuf]

Predictive Analytics in Healthcare

Predictive analytics is a field of study that consist of using statistical technique to predict future events. It’s used to predict weather forecasts, risks and opportunities in business, to predict what a customer may want to buy, to predict neural networks in the brain, and in many other areas. It uses several models to predict outcomes such as qualitative models, time-series models, and causal models. The Delphi technique, for example, is a qualitative technique that incorporates estimate made by experts anonymously, making more estimations after discussion between the experts, and then averaging the final estimations. This and a bunch of other techniques have been helpful in making predictions.

It has been proven to be predictive of events in Amazon delivery outcomes, Facebook ads, the Obama/Biden political campaign, and many others. Amazon has been able to predict when packages will reach their destination with 99% accuracy. Facebook as been successful in producing adds that are the most appealing it’s users based on their likes. Predictive analytics made Obama aware of what people were discussing online so Obama was able to speak to the people about the things that were concerning them. Even though predictive analytics have been successful in these case, it is not always used to its full potential especially in healthcare.

Predictive analytics uses several stratification methods to help improve patient care. For example, elder risk assessments uses patients’ ages, genders, marital statuses, number of hospital days, and the medical illnesses that they have in order to categorize the patients into specific categories. The category a patient in is appoints them to a specific routine of care. In addition, the Charlson Comorbidity measure predicts the rick os one-year mortality for patients depending on hat illness they have. The predictive methods, however, do not include the patient outcomes of their care in their models which is needed to provide more accurate predictions such as hospital readmission.

Furthermore, many technology advances have been made in healthcare such as in DNA sequencing, gene editing, and others. However, using the data produced from these technologies can’t be utilized fully to make predictions because of the context of when the data was produced. One person’s DNA sequence can change between the time is was sequenced to the time it is use for analysis. Healthcare systems may also lack the resources to utilize this data fully. Knowing the outcomes of patient care and having the right resources can help solve this.

To make this data more useful in predictive analytics, a health care system must have the best resources available and be aware of some guidelines. These include using tool such as Weka, Orange, and R can make make accurate predictions from big data. They should also keep in mind the context of when the data was taken, to include outcome data, and to make sure that predictions are content driven and clinician driven. The content should be utilized fully and clinicians should act on the predictions made. 

Sources:

https://channels.theinnovationenterprise.com/articles/80-top-5-analytics-success-stories

https://www.healthcatalyst.com/predictive-analytics-healthcare-technology

https://www.healthcatalyst.com/3-reasons-why-comparative-analytics-predictive-analytics-and-nlp-wont-solve-healthcares-problems/

https://www.healthcatalyst.com/understanding-risk-stratification-comorbidities/

https://www.healthcatalyst.com/predictive-analytics-healthcare-lessons/

[tobiyusuf]

Privacy and Security in Healthcare

A lot of data in healthcare need to be protected and private from outside parties. Several rules need to be provided when it come to sharing, protecting, and keeping information from patients private. These types of information need to be protected by following the Health Insurance Portability and Accountability Act (HIPAA). 

The type of information that need to be confidential are patient’s history, their condition, treatment, and demographics. Patient’s also have to right to chose how they want their information to be kept private. This information can only to disclosed to the individual and health care providers who need to know a patients condition in order to continue caring for the patient. However, healthcare providers also are not allowed to share the information they receive from a patients because this will violated HIPPA rules. When working as an EMT, I had access to patient information because this information needed to be recorded and passed on through transitions of care from one health provider to another. I also needed to refrain from sharing this information to people outside of the line of care. 

So how is this information kept private? Well, with the transition of paper records to electronic health records, more measures need to be take to keep data more secured. These records can be in danger of getting hacked, lost accidentally, stolen, or destroyed. Several ways that information is protected are through pins and passwords, encryption, and others. These methods however may not be enough to secure these data, so more levels of security needs to take place. For example, I had to use duo factor identification during my last Co-op at a research institution. It consisted of logging in normally with a username and password and then also having to ensure that you are the one logging in by enter a pin on another device like a cell phone, by receiving a call, or by another method. In addition, computers need to be protected from malware such as viruses, trojan horses, spyware, etc. these can be protected from antivirus software but one needs to careful when installing one and keep it updated. These software may be viruses themselves and cannot be trusted. 

Encryption is another widely used method of security use in healthcare environments. Encryption is a good way to protect data it prevents third parties access to information without a decryption key. For example, I worked in a health care environment where I had to write recalls for patient to reschedule more appointments in the future. To do so, I had access to patients notes from their appointments because the computer I used as a decryption key that allowed me to read their information. however, there were some cases where I still could not get access to some patient’s information because some patients request for  more protection of their information than others. I had to put in another key in order to get access to it. Encryption allows for several levels of security.

The problem with encryption is that it can under used since the laws of using encryption are vague. This can cause employees to not use it properly and increase risk of exposure of information. Research show that the number one reason for security breaches are from employee negligence. More emphasis needs to be taken when securing data within the work place.

As I move forward in my career, I will make sure to keep updated wit the more advanced methods taking place to secure information and be aware of what I do when i get access to private information. Keeping patient data private is important since it is a human right for patients to to get access to it and keep it private.

Sources:

https://healthinformatics.uic.edu/resources/articles/confidentiality-privacy-and-security-of-health-information-balancing-interests/

https://www.healthit.gov/buzz-blog/privacy-and-security-of-ehrs/privacy-security-electronic-health-records/

https://www.apricorn.com/press/data-encryption-in-healthcare

[tobiyusuf]

Information Processing in Computers and the Human Mind

Since I am interested both in how computers deal with information and how the human mind deals with information, I decided to write about the similarities and differences on how computers and human minds process information. 

A computer codes information, stores information, uses the information, and produces an output. Information gets stored and organized in a database. A type of database are relational databases that organize data in connected two-dimensional tables. A Database management system provide the means for creating, maintaining, and using the databases. Certain applications will request data from a database so the database sends them the data. The application in turn will update the data and send it back to the database. Certain languages such as SQL or C++ are used by applications to retrieve the data and translate it. Data gets processed in the application. Data is not only stored in one database but in many databases which can be difficult for applications because they may have to retrieve information about one customer from different databases. This can be tiring to do. Because of this, databases need to communicate with another and they do so through another language such as XML. Things get more complicated in that one application may not know how to read an XML file sent to them from a database, so another language need to translate this information to a language that the application can understand. XSLT is used to do this. This process can get very complicated and confusing for some but the human mind gets even more complicated.

So how do humans process information? Human brains do code information store information, and produce an output like computers. In the brain, information is pretty much stored all over the place. We retrieve information from all of our senses: sight, smell, touch, taste, audio, and others. For example, a person can hear something which is sent through sound waves through the ear. The hair cells code this information into neural activity and is then stored further along in deeper areas in the brain. If the sound ended up being to loud, the brain may elicit an output response which will cause the person to cover their ears. There are many other areas such as the temporal lobe that processes visual information and the amygdala that process our emotions such as fear or anger. Other tasks that the brain must do may need to get access to information from many areas as a time. This is similar to databases in that one database may only contain one type of information while another one will hold a different type of information. Different companies with different goals in mind may only need refer to one database where others may need to refer to more. Additionally, just like how a databased has a limited storage of information, a brain also does. Humans have selective attention where they only hold on to information that is meaningful to them. 

Furthermore, the information process may be different between computers and human. computers may only be able to process one task at a time and must be completed before going unto the next one which the brain can do many things at once.

Overall, processing information gets very complex in both computers and in humans. while researchers are trying too make this process more organizes in computers, the way we process information cannot possibly get any easier. However, certain diseases of the brain can disrupt how we process information and this is something that I am interested in learning more about.

Sources:

http://www.dataversity.net/brief-history-database-management/

https://www.simplypsychology.org/information-processing.html

[tobiyusuf]

Excel and Essential R

When it comes to spreadsheets and statistical analysis, most people with think of using Microsoft Excel and not many will think of Essential R. This makes sense because I grew up using excel when I had to do homework assignments such as lab reports. It’s not very hard to use when trying to record put simple data and make graphs and charts. However, Essential R is another statistic tool that can do much more than excel.

Essential R is better than Excel in several ways. R can handles more rows and columns than excel. Excel can still hold a lot of tables and documents but overtime, rows and column be created more and more by several different people. R can also calculated data faster than Excel and can handle a lot of data points better than Excel. It is easier to also used R source codes on different databases. You can easier attach different databases on R and then apply the readily defined functions of R on each database easily. It excel, it may be more troublesome to do so. You may have to change from one sheet to another to go to different databases. Also, applying functions to Excel between databases can be time consuming. In addition, R has libraries that anyone can get access to for free. Excel is not a free program and is not very sharable. R has a very active community who improves and increases functions in R. Lastly, creating graphs and data visualization in R is easier to make and to understand.  

It may seem that Essential R is the best program to use, I would choose Excel over R in several cases. If I am dealing with only a small amount of data from a lab experiment, I would use Excel since it isn’t too complicated to deal with small data on Excel. I also prefer having each row and column to be separated by lines. Essential R does not have this type of separation. I’m also very used to using Excel and would probably continue to use it more than R. I would recommend using R if you know that you will be working with in between a lot of databases and want to do more complicates statistical analysis.

If I were to continue to work in a health care environment, I may prefer to use Essential R when analyzing research on many different topics. Health care systems contain holds a lot of records a lot of complex data which would be harder to interpret if I were to use Excel. I will keep this in mind in the future.

Sources:

https://www.gapintelligence.com/blog/2016/understanding-r-programming-over-excel-for-data-analysis

http://georgejmount.com/r-vs-excel-vlookup-vs-indexmatch-enough-with-the-false-dichotomies/

[tobiyusuf]

Relationship Models in Healthcare

Relationship models are the most commonly used data structure today. They come in the forms of tables with rows of entrees and columns that contain attributes of the entries. Each row or entry in a data record contain unique attributes and have a primary key. A database can contain many tables that are related to each other through foreign keys. Rows, columns, and tables can be retrieve through selection of rows, projections of columns, joins of rows of different tables, and unions. It is important that there are no duplicate entries or redundancies in these tables. The models exist almost everywhere in modern businesses.

Data in health care however may be to difficult to be distributed in relationship models. Simple data such as the name, birthdate, address, and phone numbers of a patient or doctor can be simple to organize and retrieve. But, what if there are more than one person who has the same first name and last name, what if more than one family member has the same phone number? Also, there are more complicated relationships that occur in a health care database. A physician may work in more than on department which will require more foreign keys to be liked between the physicians table and the department tables. A patient may have more than one disease and take more than one medication which can make it hard to distinguish which medication is taken for which disease. Also, certain disease may share the same symptoms which can cause to misdiagnosis. Models maybe geared towards patients but not towards health providers and vice versa. This data becomes more and more complex.

Comparative Effectiveness Research (CER) is being done to make data more organized and more effective in guiding decisions. The Scalable Architecture for Federated Translational Inquiries Network (SAFTI Net) aims to do this. It provides a case study that describes ways to prevent diagnosis. It combines clinical and financial data from electronic health records into a analytical-only database in order to answers question about the effectiveness of treatments, diagnostics and protocols. It adapts an existing model that contains some data that is not suitable for CER and data that was suitable in order to see if effective decisions can be made from a complex data models. The SAFTINet thought of criteria necessary for data models to provide effective data. The Observation Medical Outcome Partnership (OMOP) data model contains these criteria. It allows for a broader range of clinical observations without changes to the model, handles missing data without creating empty cells, and supports complex analytic methods.

Health care providers must be aware of how complex health care data may be. retrieving this data may not provide the correct decisions to be made when caring for a patient. Databases need to improve so that they can proved effective decision making data. The OMOP database in one step in improving this issue. I would like to take part in improving health care through this process.

Sources:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3824370/

https://www.healthcatalyst.com/healthcare-database-purposes-strengths-weaknesses

[tobiyusuf]

Healthcare Databases

Modern businesses have databases that contain tons of interrelated information. The move from paper records to digital data is becoming more prevalent. Databases are built for a specific purpose, not just for storing information. Their purpose is to allow for easy storage, access, and retrieval of information for a specific purpose according to its users. It is structured in a way most beneficial to its intended purpose. Members of a business can retrieve information is a much faster pace than before.

The most commonly used database in health is OLTP (online transaction processing). Several computer applications runs on the OLTP database such as EHR (electronic health record) which allows one to instantly find information about a patients, lab systems, patient identification, research, and much more. These databases allow for standardization of the many processes that exist in healthcare and information can be backed up and secured. OLTP also allows for most cost effective care because of the large amount of data it holds.

Databases may seem to be the most effective ways to store, access, and retrieve information but this may not always be the case. Having access to all of this information may be overwhelming for some people. People won’t be able to retain it and gain knowledge of it. Information needs to turned into actionable knowledge, targeted to solving issues rather than just being available. Also, having this data might prevent patient-physician relationships. Patients may be seen as only “data points” and not human”. Data such as signs and symptoms may not be enough information to hale  treat a patients and other information that cannot be put into numbers may be necessary. In addition, each application has its own OLTP database which makes the data from each application become isolated from the others. This becomes an issue for trying to incorporated data from different data to gain knowledge and insight about a topic. A solution for this an OPAL (online analytical processing) database. 

OLAP is a type of warehouse which basically is a layer on top off all of the OLAP databases which allows for more meaningful analysis on all of the data from each OLTP databases.

When it comes to data, one must be aware that access to all of this data may not be beneficial for administrating effective care for patients. There is no doubt that care had been improved because of the easy access to information, but the data needs to be targeted more to specific issues, and be accessed from all applications that store the data. I hope to keep this in mind in the future if I continue to work in a healthcare setting.

Links:

https://www.healthcatalyst.com/healthcare-database-purposes-strengths-weaknesses

https://www.healthcatalyst.com/healthcare-data-help-not-hinder-human-endeavor-has-2014

[tobiyusuf]

Representing and Organizing data in Health Care

So much data gets recorded and represented in a Health Care environment. It can come in the form of patient records, patient care reports, patient recalls, supply inventories, staff directories, patient directories, hospital records, and many more. All of this information is organized in some type of ontology. It can then be put into a computer and represented in a certain way such as in classes or hierarchies. Programs can be written on this data in order to provide more information and knowledge. This post will talk about ways health care data is represented how complex it in to organize this data.

When people think about how data is organized, they usually think about hierarchies, taxonomies, partonomies, and relationships between objects. For example, an topic discussed in health care are Drugs. Underneath the overarching topic can be types of drugs such as stimulants, depressants, and hallucinogens. Different drugs can be placed in each category. Drugs like nicotine and caffeine can be under the stimulant category, alcohol and marijuana can be under depressants, and LSD and marijuana can be under the hallucinogen category. Some drugs can be under multiple categories such as marijuana which is both an hallucinogen and a depressant. Also, all drugs under each of the categories are related to each other because they have similar affects on humans. The types of representation of data seems simple and straight forward but all data in healthcare is very complex and cannot be organized in this simple way.

Health care data is so complex because it comes from many different places. for example, data about one single patient can be located in different health care facilities. This is because a patient can receive care from more than one place. He/she may have been cared for in an ambulance and then transferred to an ER in a hospital, and then transferred to a rehab center. Also, medical instruments such as fitness monitors and machines that monitor patient’s vital signs record data in different ways. In addition, different departments in a hospital such as radiology and pharmacy can have data about the patient. Different healthcare providers document data in different ways than others. One ambulance service might record patients data with another system that follows one structure while another ambulance service can use another. All of these characteristics makes it difficult to standardize data and manage it.

An example of a program that aims to better organize patient data is one called Epic. It is used by patients, health care professionals, hospital staff and volunteers alike. Patients are allowed to speak with their doctors and make appointments, patient’s demographics, personal information, medical history, meeting summaries, and more are located in the software. It makes it easy for hospital staff to learn about a patients, call patients for appointment and scheduling reminders, and do much more. 

As someone who uses Epic, I’m amazed on how much data is stored in it and how a lot of tasks can be done from the program. However, I find it hard to navigate the software. I want to think of ways to better represent and organize data  in the health care system. I hope that my efforts can allow efficient care for all patients no matter how complex the data from the patients may be. The Health care data is only going to get more complex so programs like Epic need to improve as more data is created.

Sites:

https://www.healthcatalyst.com/5-reasons-healthcare-data-is-difficult-to-measure

http://www.epic.com/

[tobiyusuf]

Knowledge and Cognitive Bias

[tobiyusuf]

About Myself

Welcome! I’m a Nigerian-American Muslim nerd girl (wow that was a lot). I’m a pre-med student majoring in Behavioral Neuroscience and minoring in Information Sciences. I also enjoy musical theatre, learning about different cultures, music, art, and sci-fi/fantasy books and TV shows. My page will consist of blog post about my aspirations and goals that I hope to achieve in life. As of know, I want to become a doctor in the future and to also improve how data flows in a the health care environment. I also care about the need for more diversity in the health care environment. I may also include some posts and re-blog about nerdy things just because I can. I hope you enjoy my post that will come!