Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Feel free to add your own ideas here. If you prefer, you can send us a message with your ideas here
We can't promise that developers will choose to work on every suggestion, but if you're coming along to the hackday, we've scheduled some time on Monday evening and Tuesday morning for pitching projects so you can get people excited about your idea.
An alternative to postal questionnaires for standard assessments
Proposed By: Dr Maria Jalmbrant, Senior Clinical Psychologist at the Lewisham adult ASD & ADHD service, Lewisham University Hospital
This is an issue that if solved would have massive impact on our (and probably others') services as well.
In keeping with the requirements for evidence based practice, we use a range of screening and outcome measures. These are questionnaires that we ask patients to complete or varying formats (typically semantic differential or Likert scales). We are currently using paper copies that we are posting out and requesting that our patients return by post, or complete in sessions. When these have been returned, we spend a large amount of our clinical time scoring these and creating data bases of these scores. I have tried to search for a better way of doing this – ideally electronically since it was a while now since we entered the 21st century, however I have not yet been able to find a suitable programme where the questionnaires can be inputted and subsequently scored on an individual case by case basis according to the varying scoring rules for each scale. Most programmes (e.g. survey monkey are geared towards groups and is able to summarise scores by calculating means and other measurements of central tendency, but this would not be of much use in this context. In an ideal world, we should be able to email a link to our patients, ask them to complete the questionnaires online or ask them to complete the questionnaires in the waiting rooms of the clinic on handheld devices, immediately be able to access their summary score as well as individual responses on a computer programme that is acceptable to the trust computers, and also be able to access the scores in a word file (the patients individual summary scores to be uploaded on the trust patient data system - PJS) and an excel file (for summary data analysis of some or all patients).
Better ways of handling mental health data in electronic health records
Proposed By: Richard Jackson, Text-Mining Lead, Clinical Informatics Group
Despite the existence of structure fields in the SLaM clinical record, much of the useful information is entered as free text. In psychiatry, probably more than in any other area of medicine, it is difficult for clinicians to accurately express their observations about a patient using dropdown menus and standard coding schemes or taxonomies like ICD10 and Snomed CT. If we wish to do any computation on this data, we currently have to use natural language processing tools to extract the information from the text. This is challenging and inevitably not 100% accurate. If we accept that a very structured EHR isn't going to work for psychiatry, then maybe we should just try to get better at capturing the data in free text. We would like to try to build an interface that lets clinicians enter their observations and reports in free text, but also helps them to use standard terms, flags data that really ought to go in structured fields like measurements or test results etc. But, this assistance must be un-intrusive and fit in with their workflows. If it all goes a bit Windows Paperclip they'll just turn it off.
Patient Record Timeline
Proposed By: Richard Jackson, Text-Mining Lead, Clinical Informatics Group & Prof. Rob Stewart, academic lead for the Clinical Record Interactive Search tool.
One of the biggest issues clinicians have with the move to an electronic health record system at SLaM is that it is no longer possible to pick up a patient's record and flick through the documents in it, to get an overview of their case, or of recent events, or just to find a document you're looking for. Instead, in the existing system one must open up documents one by one and read through them, which is time consuming. It would be very useful to have a scrollable timeline of all of the documents, potentially just one long concatenated document, or alternatively something that made it easy to get a summary / preview of each document as you scrolled, without having to download it and open in on your local machine. In general, better ways of visualising the key points of a patient's record and drilling down to specifics in a user-friendly way would be extremely useful.
Rob Stewart can't make it to the hackday, so he's provided a detailed description of the problem to guide any team that works on this problem: Clinical Timeline Details
Information extraction / NLP from the text component of the clinical record
Proposed By: Cass Johnston
Lots of psychiatric data is entered into the clinical record as free text. We have some dummy data. If you have any experience with information extraction / NLP tools there are lots of potentially useful bits of information you could have a look at.
Agent-based alerting Systems for Clinical Records
Proposed By: Richard Dobson / Zina Ibrahim
We have access to a development version of SLaM's Electronic Health Record, with a small amount of dummy data to play with. Zina will provide a quick into to JADE, a framework for building agent-based systems in Java. Agents can monitor a patient's entire record and can alert clinicians to things they might not have seen.
If you would like to participate in a JADE-based project, please have a go at getting Eclipse and JADE working on your laptop before the start of the hackday. See the JADE page for more info
Waking from Sleep App
Proposed by: Bob Patton
An App that detects waking from sleep. This has clinical utility in a number of areas, primarily in studies that require ‘waking’ data collection or sample provision (eg. Cortisol levels in sleep surveys). In this case the app would detect when awake and then display a predetermined message, link to a study website , send a reminder txt etc. If this existed we would use it in an on-going medicinal trial where community detox patients would be instructed to provide waking saliva sample for cortisol measurement
Mobile Field Impairment Test (FIT)
Proposed by: Bob Patton
A smartphone based Field Impairment Test (FIT) – this is the roadside test utilised by various law enforcement agencies to test for sobriety, and consists of a number of physical test. The app would utilise the accelerometer and proximity detector to provide a decision on sobriety. This has many practical uses - I understand that police currently have access to a dedicated handheld device that does this, but it is prohibitively expensive. From a research perspective the FIT test could part of on-going data collection or outcome measures for studies of substance related impairment.
The current FIT tests used by UK police (to be built into the App) are:
a. Examination of pupils. A subject’s pupils are examined and if they are outside the normal range of between 3.0 and 6.5mm this is recorded as abnormal. b. Romberg test. The Romberg test is a test of the subject’s internal clock. The subject is asked to tilt their head back slightly, close their eyes and estimate the passage of thirty seconds. Results of between 25 and 35 seconds are normal. c. Walk and turn test. The subject is asked to stand with their right foot in front of the left foot, touching heel to toe. They are asked to take nine steps along the line, turn and walk nine steps back. The subject must count out loud and look at their feet while doing the test. If the subject fails to count out loud, look at their feet, loses balance etc, these failures are recorded. d. One leg stand test. The subject is asked to stand on one leg with the foot raised 6 to 8 inches (15-20 cm) parallel to the ground. The subject is told to look at their foot and count out loud. e. Finger to nose test. The subject is asked to extend the index fingers of both hands and hold them palms facing forward. With the head tilted slightly backwards and eyes closed the subject is asked to touch the tip of the nose with the tip of their finger with the hand indicated by the officer.