Changelog-Engine API

API 3.20.1 is now live!

This was a maintenance update where we:

• Made backend improvements that focused on the scalability of our API.

API 3.20.0 has just been released!

With this update, we made:

• Technical enhancements to how the algorithm chooses what question to ask next, which will allow us to make even more qualitative improvements to this process in the near future.
• Technical enhancement to Red Flags functionality, making it more independent from the logic associated with next question selection, which will enable us to develop each more effectively.
• Improvements to Search endpoint data logging, allowing us to gain a better understanding of how patients use this functionality.

Changes from this update will not impact the end results of either the triage assignment or the condition rankings.

API 3.19.0 released!

With it, we made:

• Technical enhancements to Rationale and Explain functionality - We’ve refactored the parts of the Engine that are responsible for providing data to the /rationale and /explain endpoints, improving maintenance and allowing for more future improvements of those functionalities.
• Fixes to a issue related to the reporting of Age=0 years - We’ve discovered an infrequent situation where some applications would report a patient’s age as 0 years but not provide the exact number in months. As it is medically important to specify an infant’s actual age, we’ve prohibited the providing of age value = 0 years. In such cases, an error message will be returned. To avoid errors, ensure that anyone under the age of 1 has their age reported in months.

API 3.18.0 live now!

In this update:

• We removed Lab tests from the /search endpoint,
• We removed Lab tests from the /concepts endpoint,
• We removed the Lab tests count from the /info endpoint.

API 3.17.1 released today!

Here are the latest improvements:

• We've ensured that prevalence values are always returned in lowercase,
• We've removed the prevalence value LIKELY.

API 3.17 is here!

We'd like to introduce API 3.17, which brings key enhancements like:

• Improvements in logic to the Red flags method in Suggest - We’ve evaluated and fixed several issues related to the Red flags method in /suggest, which should improve the quality of outcomes from that endpoint,
• Fixes to Evidence-based risk factors method in Suggest - We’ve introduced a fix to the endpoint that should increase the number of relevant risk factors it returns,
• New additions to our API - We’ve added a new parameter to /search. Now you can use the include_pro parameter to decide whether users should be able to search for more complex medical symptoms, such as “Moniz sign” or “Trigeminal neuralgia”,
• Technical enhancements to Suggest’s functionality - We’ve introduced changes across the whole of Suggest’s logic that make it easier to maintain long term while also not impacting its outcomes.

API 3.16 now live!

We'd like to introduce API 3.16, which brings key enhancements like:

• content filters for improved search functionality - new content filters were added internally to ensure more refined search results, enhancing search accuracy and improving the user experience,
• /suggest and /diagnosis endpoint enhancements - multiple updates were made to improve the long-term maintainability and performance of the API and Inference Engine,
• teleconsultation_applicable flag improvements - despite this flag being deprecated, continued support and improvements are being made to its functionality to ensure that its users continue to have seamless integration.

API 3.15 is live!

With it, we:

• introduced several changes that’ll improve our ability to maintain the API and Inference Engine long term,
• updated the question selection algorithm by removing some edge cases and simplifying the overall logic of choosing the next question,
• solved a bug related to symptom duration recognition in our NLP,
• deprecated the teleconsultation_applicable flag in the /triage endpoint. This functionality can now be achieved by using the /recommend_specialist endpoint. For more information, see Specialist & Channel Recommender.

API 3.14 is here!

We're thrilled to introduce API 3.14, which brings significant improvements to our platform like:

• an improved triage algorithm - our triage algorithm has been fine-tuned to minimize over-triage during emergencies, leading to more accurate assessments and reduced unnecessary alerts,
• streamlined maintenance - we've implemented technical improvements that make maintaining our API and engine easier.

API 3.13.1 is now live

Here's what's new!

With this update, we made a couple of minor changes to our newly released Intent Survey feature. More information about the feature can be found here.

We've also added three new Asiatic languages for our API customers: Thai, Japanese, and Chinese Traditional (Hong Kong). This brings our number of available languages up to 24.

API 3.12 has just been released!

Here are the latest changes:

• We introduced several changes that’ll improve our ability to maintain the API and Inference Engine long term.
• We fixed a bug involving automatically inferred risk factors.
• We fixed a bug related to the reporting of non-existent evidence IDs.

API 3.11 is now live!

Here are the newest changes to our API:

• We introduced several security improvements to our core functionalities.
• We introduced changes that will improve the maintenance of the API and Inference Engine long term.
• We fixed a bug related to the impact evidence was having on inference.

API 3.10

API 3.10 is here!

We are introducing a new enable_symptom_duration feature. This feature will allow you to enable an additional parameter that will let you specify the exact duration of the reported symptoms.

For details, please check Diagnosis v3 - Infermedica for Developers

API 3.9 is here!

With it we are introducing internal security improvements that are making the user data and interview process even more resilient.

API 3.8

API 3.8 is here, along with a new feature!

Discover the Patient Education feature that is now available with API 3.8. This new feature brings an additional layer of knowledge, in the form of articles describing the most popular conditions and some common care methods.

The Patient Education functionality involves changes in two endpoints:

• /diagnosis (endpoint extension)

  • When extra "include_condition_details": true is applied, new information has_patient_education will appear to every condition in ranking.
  • Diagnosis response is extended with a new field interview_token which is needed to use the Patient Education functionality.

• /patient_education (new endpoint)

  • Endpoint returns Patient Education content in JSON rich text format.

Apart from the introduction of Patient Education, API 3.8 also brings a few internal security improvements.

Please note that the Patient Education feature is available for an additional cost. For details, please contact the Infermedica team.

API 3.7

API 3.7 is here! Below are the main changes:

• To improve compatibility of /triage and evidence seriousness levels we added a new field: seriousness. Possible values for the field are: serious, emergency and emergency_ambulance.
• We deprecated flag is_emergency for items in the serious evidence list. The evidence seriousness (value of emergency and emergency_ambulance) is more granular and should be used instead. More details can be found in the Triage section.
• We applied internal security improvements.

Engine & NLP

The latest change to our Engine & NLP:

• We overhauled the spelling corrector used by our NLP endpoint. The new algorithm is capable of dealing with common mistakes going beyond typical typos. Thanks to it, the /parse endpoint will be able to understand messages such as “soar throat”, “diaheria” or “backpain”.

API 3.6

Long-awaited update to Infermedica API just arrived! With API 3.6 we updated third-person questions in /symptoms and /risk_factors. More details can be found in the Medical concepts section.

API 3.5

API 3.5 is here! Below are two main changes:

• We introduced new suggest modes for risk factors based on provided evidence. evidence_based_risk_factors gives you an option to ask our system if we have propositions of any risk factors based on the already collected evidence. demographic_risk_factors returns risk factors that are common for given age and sex. demographic_risk_factors works in the same way as the current method risk_factors and we will support both for the time being. More details can be found in the Suggest related concepts section.
• We adjusted the text form for explanations (explication and instructions in /diagnosis) to consider patient sex.

API 3.4

Below are the newest changes to our API:

• We extended concept types within /search with condition. You can now use /search to request a list of all available conditions.
• We added the possibility to display explanations (explication and instruction) for returned items within /suggest. You can find guidelines on how to use /search effectively in the Suggest related concepts section.
• We added a possibility to disable asking about intimate concepts within /diagnosis and /search. More details can be found in the the Suggest related concepts and Diagnosis sections.
• We are now filtering out items that are difficult to understand within /search. This will prevent us from displaying symptoms which cannot be clearly described by its common name without asking a dedicated question.

API 3.3

Below are the newest changes to our API:

• We added a new field in the /triage response - root_cause. Now, understanding the reason behind triage recommendation is much easier. See more details in Root cause levels paragraph within the Triage section.
• We strongly believe that the importance of a quick response in emergency cases is unmeasurable. This is why we added the new has_emergency_evidence flag in /diagnosis. This flag informs about the high seriousness of reported evidence. It is a helper for the purpose interrupting interviews in emergency cases. See more details in Stop recommendation paragraph within the Diagnosis section.
• We improved the algorithm of suggest for red_flags to better detect most common emergency cases.

API 3.2

February 2021 welcomes us with new changes. Our team hasn't slown down and we've made some technical improvements to our API. We've laid the groundwork for our future updates but, most importantly, we refreshed our suggest algorithms. The new implementation of /suggest endpoint in symptoms mode will cover a wider range of cases, as well as make it faster and more accurate. The changes do not require any actions from the API user for now, but stay tuned!

API 3.1

November is almost over, and we introduced some significant changes to our platform. One of them is Specialist Recommender, a new powerful feature, enabling healthcare providers to enrich digital pre-diagnosis and triage processes.

• We added an endpoint for recommending the most relevant specialty of the physician and appropriate consultation channels. See more details in Specialist recommender section.
• We added a new mode in /suggest endpoint to return common risk factors given patient demographics. See more details in Relevant risk factors paragraph within the Suggest related concepts section.
• The red flags can be now used alternatively in /suggest endpoint with {"suggest_method": "red_flags"}. We recommend using the /suggest endpoint, since /red_flags has been deprecated. See more details in Red flags paragraph within the Suggest related concepts section.
• More condition details in /diagnosis are returned with an extra parameter include_condition_details. See more details in Include condition details section.
• The “common_name” field in /symptoms and /risk_factors is not null anymore.
• Endpoint /search returns correctly filtered concepts in misspelling.

API 3.0

In API v3, we introduced a few significant improvements that require the adaptation of your application.

• We significantly improved the reasoning algorithm. It outperforms the previous algorithm in terms of diagnostic and triage accuracy. Now, a more relevant ranking is returned.
• We added the age dimension to our models, which requires a new request format. Moreover, the age attribute becomes mandatory in /parse, /suggest and GET requests. Additionally sex parameter will be mandatory in /suggest.
• Requests with age under twelve years old will not be accepted anymore.
• A new /concepts endpoint was introduced to return basic data about symptoms, risk factors, lab tests beta
  Lab tests are still in development and will be available soon.
  , and conditions. See more details in the Concepts section.
• A new option in the extras collection: enable_explanations was introduced in /diagnosis. The purpose of this option is to provide additional descriptions of selected questions to make them easier to understand for users. See more details in the enable_explanation section.
• A new option in the extras collection: enable_third_person_questions. This option allows you to create an interview scenario in which users can answer questions on behalf of someone else, e.g. "Does she have a headache?" See more details in the enable_third_person_question section.
• The endpoint /rationale will return an empty response when there is no result for the request. Previously it returned a 400 error code and an error message. See more details in the Rationale section.
• 5-level triage recommendation will become the default one. If you’d like to continue using 3-level triage, you need to add the extras option: enable_triage_3: true See more details in the enable_triage_3 section.
• Diagnoses ranking is limited to the most relevant conditions by default. The extras option enable_adaptive_ranking: true is no longer supported. See more details in the Disable Adaptive Ranking section.
• Flags initial and related that occur with evidence in POST requests are no longer supported. They are replaced by the source attribute. See more details in the Indicating Evidence Source section.
• Request parameter type in /search is replaced with parameter types which can handle multiple values separated by a comma.
• Endpoint /lookup is removed.

API 2.5

Most of the changes were related to technical and infrastructure improvements.

• We shortened the maximal length of the interview in /disable_groups mode.
• Triage “consultation” is now recommended in case no diagnosis was found.
• Since this version, you could check which API version is already deployed via /info endpoint.

API 2.4

The last two months were productive in our team. Check the new update - it's a breaking change!

• [BREAKING CHANGE] We introduced an evidence attribute called source which allows you to describe the exact stage of the interview at which given evidence was reported. Attributes initial and related get deprecated. Read more in Indicating evidence source.
• With new extra parameter enable_adaptive_ranking, you are able to limit diagnostic ranking to most relevant diagnoses worth displaying to the user. Read more in Adaptive ranking.
• Last but not least, the endpoint /suggest is fixed, so you are able to filter desirable results accurately.

API

• Our NLP service (/parse endpoint) now accepts text input that is 2048-characters long. This may help you better understand your users’ health concerns.
• We’ve released a new version of our Python client (opens in a new tab) that supports /red_flags and /rationale enpoints.

API

We've released a new version of our triage algorithm, which in addition to the ranking of possible conditions, puts more emphasis on the symptoms reported by the patient. This will also allow the upcoming content release to include new, more fine-grained triage levels for more meaningful recommendations.

API

• Now, question selection algorithm prioritizes risk factors (e.g. being post-menopausal). As a result, some conditions are excluded.
• We’ve improved stop recommendation algorithm.
• We’ve fixed bugs that prevented mentions of some common symptoms from being properly captured by /parse.
• Lastly, we're working on a German language support for /parse. Beta version will be available soon!

API

We're happy to share two new features with you.

• Have you ever wondered when is the good moment to stop asking further diagnostic questions? We've introduced our stop condition recommendation that you can use to end interviews when enough information is collected.
• Our question selection algorithm can now better focus on the initial symptoms reported by a user. All chief complaints and reported observations should now be marked as initial evidence. This is an important change that you should implement for the optimal performance.

API

We continue to improve already existing features:

• Question selection algorithm – Enhanced for risk factors related to injuries.
• /search endpoint – We've rewritten phrase matching algorithm. The endpoint now returns more accurate results, especially in Chinese (due to better tokenization).
• /suggest endpoint – We analyzed our statistics once again and, as a result, recommendations have become better.

API

We are introducing common names, one new endpoint and more:

• The API now returns both name (professional medical term) and new common_name (plain language) – for most of the symptoms and many conditions.
• The new /suggest endpoint returns related symptoms. It allows you to simplify the initial data entry by suggesting other symptoms that patient may have.
• The /search endpoint was improved to work better with languages using diacritical marks (e.g. Spanish).
• NLP understands common risk factors.

API

We profiled crucial parts of our code and achieved meaningful performance improvements. A significant focus has been put into improving question selection strategy for conditions strongly associated with a particular risk factor (e.g. skeletal trauma).

API

We've introduced important enhancements to our underlying reasoning engine:

• The algorithm may now ask about risk factors such as chronic conditions (e.g. diabetes), events (e.g. injuries) and activities (e.g. mushroom consumption).
• Improved sequencing of questions to better differentiate between conditions.
• Improved triage recommendations as they were initially too restrictive.

API

We're exited to introducte the new endpoint. It's called /triage and it allows you to assess the severity level of reported symptoms. We published a brand new section in the documentation describing all details.

We're excited to begin this year with a major update of the Infermedica API.

API

Improvements of the reasoning algorithm and request processing.

• More flexible question selection which is now less depended on pre-fixed logic.
• Better filtering of likely conditions.
• Solved a bug causing null questions to be returned on rare occassions.
• Better error handling.
• Introduced various techniques to prevent data-scraping (malicious actions against the API are obviously prohibited).
• More scalable and reliable server infrastructre in two fail-over locations.

We’re constantly improving. From now on we’re going to post all changes here. All updates are divided into two groups:

• Model – all changes affecting the medical concepts (conditions, symptoms, risk factors and lab testsbeta
  Lab tests are still in development and will be available soon.
  ). We’ll let you know what has been added, deleted or updated.
• API – all amendments in the Inference Engine and other algorithms.

Check out what’s new and update your application’s code, if necessary.