<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Pascal Werner on Medium]]></title>
        <description><![CDATA[Stories by Pascal Werner on Medium]]></description>
        <link>https://medium.com/@pascalwerner?source=rss-e144ce40a263------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*2XRuGXuwLMUPfbjs.</url>
            <title>Stories by Pascal Werner on Medium</title>
            <link>https://medium.com/@pascalwerner?source=rss-e144ce40a263------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 23:44:04 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@pascalwerner/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[MDR for DiGA manufacturers]]></title>
            <link>https://pascalwerner.medium.com/mdr-for-diga-manufacturers-aa1c61dda279?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/aa1c61dda279</guid>
            <category><![CDATA[medical-devices]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Sun, 31 Jan 2021 16:56:11 GMT</pubDate>
            <atom:updated>2021-01-31T16:56:11.034Z</atom:updated>
            <content:encoded><![CDATA[<p>This article may be a bit different than my usual ones as it’s a summary of <a href="https://www.eventbrite.com/e/diga-360-3-austausch-infoplattform-zu-digit-gesundheitsanwendungen-registrierung-136254337391">a recent presentation</a> on Jan 25, 2021. While generally a good practices, providing my slides to the attendees seemed quite useless to me, given how minimal my slides typically are. If you find this experiment of summarizing the talk in a blogpost useful, please let me know, e.g. by commenting or liking/clapping.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ponn5f4nN0ZRVR93jWf2kg.jpeg" /><figcaption>Aviation. An industry strongly associated with processes, in development of planes as well as in operations. It provides plenty of cases that can be studied for learnings for medical device manufacturers. That may be good content for another article.</figcaption></figure><p>The presentation was themed with introduction to medical device certification within the scope of DiGAs (the new reimbursement scheme in Germany, in case you’re unfamiliar with it). The second talk of the event was on data privacy/security. So at least that could be excluded from my talk. Still a huge topic. I decided to focus on a few meta topics that are relevant for MDD but even more for MDR devices (whenever I say device, I mean medical device).</p><h3>European conformity or Chinese export</h3><p>Last things first: at some point of time you’ll put the CE mark on your device. That means everything else is done, the technical documentation (TD) is created, a quality management system (QMS) setup, and the notified body certificate(s) is/are issued, if applicable. Please know the difference between the European Conformity CE mark and the similar looking Chinese Export. Well, just use it from an <a href="https://ec.europa.eu/growth/single-market/ce-marking/">official source</a>. Also, “The CE marking must be affixed visibly, legibly and indelibly […]”: please don’t hide it somewhere in the settings. That’s clearly not visible to the typical user.</p><h3>MDD vs MDR and thoughts on transition period</h3><p>Second topic was the current state of MDD vs MDR for devices already on the market or in development. Everyone has heard by now that there is a transition period under MDR (Art. 120(3) MDR). Also nearly everyone seems to have come up with the magic trick of getting around that deadline. One of the recurring ideas is to include more indications or features in the intended use and just add them to the actual device later. The argument is that it’s not a significant change because it’s already covered in the intended use. Well, no. That would be too simple. Another recurring idea is to register shell devices where the device is heavily based on content. So you may not change much in the code but add in 95% of the content later on. Well, again, no. <br>In short: whenever it seems like you’ve found a magic trick consider this to be a significant change. You can surely still do changes to the device after May 26, 2021. But that is restricted to security fixes and maintenance in general. While not every trick may get caught right away, I wouldn’t build a business on this. It will also make your MDR certification harder than necessary.<br>Talking about MDR: that’s surely a motivation to still finish a marketable device before May 26, 2021 under MDD. Simply because you MDR doesn’t just mean a few weeks or months more of work. You likely don’t have clinical data and a full QMS at that point. Also your TD may not be up to MDR standards. And the 8 months of the certification process with the notified body shouldn’t be forgotten. So, depending on your device and company setup, if you don’t manage to finish your device under MDD, you can likely work/wait another 1–2 years without a device on the market. Pre-revenue they may call it. But it’s more than that. It’s pre-anything in user’s hands. If you read this now without already having started development, you may be running out of time.</p><h3>Classification of SaMD under MDR</h3><p>Until recently it seemed that everyone is content with accepting risk class IIa (or higher) for their devices. The discussion to justify all kinds of devices as class I had faded. Well, until recently. Now several people ask again why their device wouldn’t fall class I under MDR. Some argue that their device is not different to the MDD device. Well, that’s true. The different risk class simply comes from different classification rules under MDR. So for the very same device, you will likely have different classifications for MDD and MDR, especially for SaMD/DiGA. The current general interpretation of rule 11 (for standalone software) of MDR Annex VIII is that any device that is used on patients with an existing indication/disease is classified as IIa (or higher). Only devices that are being used on humans without an indication would classify as class I, in other words devices for prevention. So if you have for example a software that helps people treat insomnia, it would likely be class IIa. While if you have a device that helps people with good sleep to maintain their good sleep, that would likely classify as I.</p><h3>Startup vs medical device manufacturer</h3><p>That being said, the main part of the talk was about the transition of a startup to being a medical device manufacturer, focus on process, and the manufacturer’s own responsibility. Being considered a medical device manufacturer instead of a startup may sound abstract. One good example that may make it more tangible is the aspect of releases. Some startups care more than others about that. Some have strict processes setup, others don’t. Especially once you enter the transition period as mentioned earlier, you will want to explicitly approve every single change/release of your device to not accidentally release something that may be considered significant. And if the device is listed as a DiGA, you have additional requirements which are actually even stricter (or at least more ambiguous). If you make a release that BfArM considers significant (<a href="https://www.bfarm.de/SharedDocs/Downloads/DE/Service/Beratungsverfahren/Pr%C3%BCfbogen_Ver%C3%A4nderungen_DiGA.pdf?__blob=publicationFile&amp;v=2">see checklist</a>) but you released it regardless, you are risking to be delisted.</p><h3>Importance of processes</h3><p>Once you prepare for a MDR class IIa process, things will change even more. your notified body will expect you to have a full QMS. That’t not necessarily a bad thing. But it surely means that the “traditional” approach of having one person in the company that “does medical device certification” is finally over. If the auditor would speak to the same person for every process, he would likely be surprised/confused. And he may ask what happens when that person is on vacation.<br>No one is expecting you to have overcomplicated lengthy processes, actually to the contrary. You should take this as a chance to align your team and setup effective processes.</p><p>No one expects that the TD a is magic byproduct or, even worse, something produced after the fact. A good device development process has the creation of the various elements of the TD built in the same way it contains the actual building. Two examples: with a bit of discipline you can setup/use Jira in a way that the user stories that you create anyways for software development would also be regarded as software requirements from an IEC 62304 perspective. And who develops software without thinking about the architecture? Yes, IEC 62304 does not require you to define a software architecture for software safety class A. Still, you likely do it anyways. And, assuming you have not just one deployable (e.g. frontend and backend), then how do you do versioning? So, you may not need an architecture by itself, but you need it to fulfil configuration management requirements. But again, it’s to your own benefit. And it’s not a lot of work.</p><h3>Responsibility of the device manufacturer</h3><p>Own responsibility. What does that mean? Basically everything. You as the manufacturer decides (1) what you consider to be a significant change, (2) what norms you consider applicable, (3) the clinical data that you consider sufficient, (4) the necessary expertise in your team, for employees and freelancers/consultants/experts, and a lot of everything else. Of course there are a lot of best practices. Please don’t start questioning whether IEC 62304 is relevant. But what approach are you following for IT security? Especially when you do something new, there may not be any best practices yet. One good examples for this is the DiGAV checklist, which covers topics such as data privacy, data security, accessibility, interoperability to name some. Some of those are pretty specific. But others are terribly vague. No one has the right answer right now. Standards are still to be established. Hopefully soon so that DiGA companies can focus on delivering great devices and establishing them in the market instead of interpreting vague requirements. But still, it’s your obligation as a company to make decisions here. Well, making decisions under uncertainty is a crucial element of entrepreneurship.</p><h3>About the author</h3><p>After having worked on medical apps in a startup (software development as well as regulatory) for various years, I currently work with various SaMD/DiGA manufacturers, mostly on regulatory compliance and product strategy. Please let me know if you have any questions or feedback on the article.</p><p>Photo by Markus Winkler on <a href="https://unsplash.com/photos/eY9oGyvvRgY">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aa1c61dda279" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Design vs development]]></title>
            <link>https://pascalwerner.medium.com/design-vs-development-5a3ca7148fd0?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/5a3ca7148fd0</guid>
            <category><![CDATA[medical-app-development]]></category>
            <category><![CDATA[medical-apps]]></category>
            <category><![CDATA[medical-devices]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Sun, 05 Apr 2020 13:19:31 GMT</pubDate>
            <atom:updated>2020-04-05T13:19:31.494Z</atom:updated>
            <content:encoded><![CDATA[<p>After writing about one of my least favourite topics (<a href="https://medium.com/@pascalwerner/medical-apps-under-mdr-all-in-or-caution-7a7d9f8d6372">legal ambiguity</a>), this article is about a more favourable topic: translating regulatory speak for startup people who work on medical apps.</p><p>This article is about a very common misunderstanding seemingly obvious terms: design and development. Well, they are obvious for medical app developers as well as for regulatory people. They just understand different things when hearing those terms.</p><p>From a medical app perspective, design typically refers to the graphical design of the app. It typically splits into UX and UI. And development means implementation into code. Some companies do strictly the first and then the second. Others have a more integrated approach.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*lKH1R5y_v_4XETIh3m1AOQ.jpeg" /></figure><p>From a regulatory perspective that all means the same. Design, development, and design and development. Design is mostly used by the FDA. Design and development is used by ISO 13485. But still, they mean the same. This also makes sense when looking at medical devices in a bit of a more general sense. Not every medical device has a graphical user interface, e.g. scalpels. Well, guess what, the FDA still refers to design when it comes to the designing and developing it. So you should understand the term design in a more general engineering sense.</p><p>One of the next articles will describe how the typical activities fit into the more structured medical device lifecycle process.</p><p>Photo by <a href="https://unsplash.com/@plushdesignstudio?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Plush Design Studio</a> on <a href="https://unsplash.com/s/photos/design?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5a3ca7148fd0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Medical apps under MDR — All-in or caution?]]></title>
            <link>https://pascalwerner.medium.com/medical-apps-under-mdr-all-in-or-caution-7a7d9f8d6372?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/7a7d9f8d6372</guid>
            <category><![CDATA[medical-devices]]></category>
            <category><![CDATA[medical-apps]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Sun, 05 Apr 2020 01:05:56 GMT</pubDate>
            <atom:updated>2020-04-05T01:07:12.267Z</atom:updated>
            <content:encoded><![CDATA[<h3>Medical apps under MDR — All-in or caution?</h3><p>Alternative title: <strong>Intention or just bad document control?</strong></p><p>Ideally this would just be a short LinkedIn post, stating that MDR gets postponed for a year. Instead things are a bit more complicated. At least for medical apps.</p><p>But when reading the <a href="https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=COM:2020:144:FIN">proposal</a> (yes, it also still needs to pass Parliament), I was surprised to not see an update to the dates in Article 120 (3). In other parts of Art. 120 and seemingly the whole MDR, May 26 2020 was replaced with May 26 2021. This means that MDR overall will be delayed by one year.</p><p>Unfortunately, 120 (3) is important for most medical apps. This is because most medical apps fall in class I under MDD and will be up-classified under MDR. For those devices, 120 (3) defines a transition period of 4 years during which you can continue to offer your MDD class I device as long as you do not do any significant changes.</p><p>This article will focus on implications of the proposal on medical apps (especially the category that is known in Germany as DiGA — digital health application).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*92vi7JX4Fb6X4r4gXulMGw.jpeg" /><figcaption>Are medical device regulatory pathways as unpredictable as gambling?</figcaption></figure><p>Just to re-iterate:</p><ul><li>MDR gets postponed, which means that the MDD is still in place until May 25, 2021</li><li>the 120 (3) transition period still starts on May 26, 2020</li></ul><p>That means there are the two options for most medical apps:</p><ol><li>pretend as if MDR wouldn’t exist. Well, until May 26, 2021. Then you’ll be up-classified and you need a notified body certificate right away. No transition period.</li><li>(semi-)freeze your device by May 26, 2020. Then you have the full 4 years of transition period, as long as you do not do significant changes. But everyone around you will still significantly change their devices.</li></ol><p>Option 1 is kind of <strong>all-in</strong>, while option 2 is the <strong>cautious</strong> one. It boils down to the question of doing significant changes during the time from May 2020 to May 2021, if you already have a device on the market. If you do not, then this year may definitely be very interesting to you. But what if you do not find a notified body? But maybe the date of the transition period gets changed. Basically punishing those who opted for 2 and benefiting those who were more risky.</p><p>Yes, these are two kind of weird options. And it doesn’t seem to be intended.</p><p>Coming back to the title and the relation to document control. MDR got updated last December with the 2nd Corrigendum (also known as Christmas present for the medical device industry). Before 120 (3) didn’t have a date included. The 2nd Corrigendum introduced dates to it. Some people believe now that the authors of this proposal were using an older version of the MDR. Well, what does ISO 13485 say about document control? Anyways, fingers crossed, this it’s being corrected. And let’s cross even more fingers that this proposal will be implemented soon.</p><p>PS.: I think that most people who are reading this article know what I’m talking about. If there are some assumptions that you feel like need more explanation, please let me know. I’m happy to elaborate on any aspect. Just don’t want the article to explode. That’s also the reason why I try to describe things in normal language as much as possible.</p><p>Photo by <a href="https://unsplash.com/@markusspiske?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Markus Spiske</a> on <a href="https://unsplash.com/s/photos/poker?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7a7d9f8d6372" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[7 questions to check the status of your medical certification]]></title>
            <link>https://pascalwerner.medium.com/7-questions-to-check-the-status-of-your-medical-certification-18b81b773a77?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/18b81b773a77</guid>
            <category><![CDATA[mhealth]]></category>
            <category><![CDATA[health-technology]]></category>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[certification]]></category>
            <category><![CDATA[medical-devices]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Thu, 01 Aug 2019 20:10:49 GMT</pubDate>
            <atom:updated>2019-08-01T20:17:39.387Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LS0zUDJNmeksF5iF" /></figure><p>This article is written for digital health startups that currently have a MDD class I product in the market or, as some people may say, a class 1 CE certification. Consider this article to be a high-level checklist to double-check your documentation. Why is this necessary you may ask? Aren’t the requirements as laid out by the Medical Device Directive (MDD) clear? Well, if you feel confident and have gone through the list of Essential Requirements yourself, you may not need to read further.</p><p>The MDD is pretty vague. And some interpretations of how to implement those requirements have been pretty low-effort in the past. While this is of course very startup-friendly, supervisory authorities may expect a different level of detail and structure.</p><p>Without further ado, here is the list:</p><ol><li>Is your documentation a large monolith? Or did you split your documentation into manageable chunks?</li><li>Do you have a clear (mental) separation between quality management system (QMS) and technical documentation (TD)? What SOPs do you have in place?</li><li>How diligent is your clinical evaluation? Have you heard of (or maybe even followed) MEDDEV 2.7/1 rev. 4? Or is it a list of some publications with some commenting text?</li><li>How does your risk management look like? Have you applied the stringent logic of ISO 14971 where you focus on medical harms? Or do you have a rather wild collection of various topics?</li><li>Do you have documented results of a market surveillance?</li><li>Would you know what to do when an incident happens that you may have to report?</li><li>Is all of that a one-man-show or a group effort?</li></ol><p>This is not meant to be a full list of what you need to have in place. It’s just a short checklist to check the maturity of your documentation.</p><p>If you happen to be a medical device professional with decades of experience, then this list probably makes you cringe. My experience of having worked with various startups in the field is that they usually have a very strong quality culture. Though just doing things right is not sufficient in the medical device world. Relevant information have to be documented.</p><p>Photo by <a href="https://unsplash.com/@glenncarstenspeters?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Glenn Carstens-Peters</a> on <a href="https://unsplash.com/search/photos/checklist?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=18b81b773a77" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Classification of medical apps under MDR — it’s not the end of the world]]></title>
            <link>https://pascalwerner.medium.com/classification-of-medical-apps-under-mdr-its-not-the-end-of-the-world-d7a1ce4b577b?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/d7a1ce4b577b</guid>
            <category><![CDATA[digital-health]]></category>
            <category><![CDATA[health-technology]]></category>
            <category><![CDATA[medical-apps]]></category>
            <category><![CDATA[gdpr]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Tue, 12 Feb 2019 20:56:29 GMT</pubDate>
            <atom:updated>2019-02-12T20:56:29.618Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>Classification of medical apps under MDR — it’s not the end of the world</strong></h3><p>Let’s not follow the crowd, but actually look into the law to understand what it means for your medical app.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pp7IY1cbqhNnHmXQyVoz8A.jpeg" /></figure><h3><strong>Short background</strong></h3><p>MDR stands for medical device regulation and is an European regulation that will become active mid of next year. It succeeds the medical device directive. And as such it’s similar to GDPR. GDPR is an European directive that succeeded the data protection directive.</p><p>If you have anything to do with medical apps you are most likely aware of that. You probably also assume that your medical app will be classified higher than it is right now. Just to remember, in Europe medical devices are generally divided into the risk classes 1, 2a, 2b, and 3. Class 1 means lowest risk and class 3 means highest risk. An Example for class 3 are pacemaker. Besides a few exceptions, most medical apps are classified as class 1 so far. The myth with the upcoming MDR is that this will change and every class 1 medical app will shift up to either class 2a or 2b. Let’s look at this in some more detail.</p><h3><strong>What does the MDR actually say?</strong></h3><p>To understand this classification a bit better, I would recommend you to look at the wording of the regulation. You find the full text <a href="https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32017R0745">here</a>. Among a lot of other things, it describes so called classification rules. You simply work through those rules and get a classification as a result. The MDR has a specific rule for software. Let’s have a look at rule 11:</p><blockquote><em>Software intended to provide information which is used to take decisions with diagnosis or therapeutic purposes is classified as class IIa, except if such decisions have an impact that may cause:</em></blockquote><blockquote><em>death or an irreversible deterioration of a person’s state of health, in which case it is in class III; or</em></blockquote><blockquote><em>a serious deterioration of a person’s state of health or a surgical intervention, in which case it is classified as class IIb.</em></blockquote><blockquote><em>Software intended to monitor physiological processes is classified as class IIa, except if it is intended for monitoring of vital physiological parameters, where the nature of variations of those parameters is such that it could result in immediate danger to the patient, in which case it is classified as class IIb.</em></blockquote><blockquote><em>All other software is classified as class I.</em></blockquote><p>Let’s dissect it a bit into 3 parts:</p><ol><li>The first part applies to your app if the app provides information that is used for diagnosis or therapeutic purposes. If that’s not the case, you can skip it.</li><li>The second part is relevant if your app monitors certain aspects of the human body, e.g. heart rate or generally any physiological processes. Again, if that’s not the case, you can skip it.</li><li>All other software is classified as class 1.</li></ol><blockquote>The main takeaway of this article is that your medical is not automatically in risk class 2 by next year.</blockquote><h3><strong>What you should do now?</strong></h3><p>A lot of the medical apps that I know do something preventative. And especially those apps won’t necessarily be classified higher than the current class 1. MDR generally gives more focus on software by having a dedicated classification rule. But that doesn’t mean that every medical app is automatically at least in risk class 2a as some people claim.</p><p>The main takeaway of this article is that your medical is not automatically in risk class 2 by next year. You need to analyze in detail your medical app and its intended purpose and decide what risk class to go with in the future. Just keep in mind, if it’s class 2a or higher, you need to work with a notified body. So you need to start this process early enough. Reach out to me if you need any guidance on that.</p><p>Photo by <a href="https://unsplash.com/photos/gM6k0JmHIBE?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">The Roaming Platypus</a> on <a href="https://unsplash.com/search/photos/classification?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d7a1ce4b577b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Digital health and its big challenge]]></title>
            <link>https://pascalwerner.medium.com/digital-health-and-its-big-challenge-8efc2d071ae4?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/8efc2d071ae4</guid>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[digital-health]]></category>
            <category><![CDATA[lean-startup]]></category>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[health-technology]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Fri, 08 Feb 2019 10:58:51 GMT</pubDate>
            <atom:updated>2019-02-08T11:02:50.433Z</atom:updated>
            <content:encoded><![CDATA[<p>Building a company is hard, every startup founder would agree with that. Having co-founded Mimi Hearing Technologies and helped building it during the last 5 years, I can very much relate to that. This article is about things that make it even harder sand why you shouldn’t worry too much about regulatory questions, including the new Medical Device Regulation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3orZX4NPEQbwNxUgz7Wm4Q.jpeg" /><figcaption>The typical tools of a doctor. The smartphone is just a fancy timekeeper so far.</figcaption></figure><h3>Why is it hard to build a digital health company?</h3><p>Building a health tech or digital health company is even harder for at least two reasons. The first reason is the regulatory side. Besides special rules in GDPR for health data, you have, depending on the risk of your product, more laws, regulations, and norms to follow. You may also need to work with a notified body to actually be allowed to put your product into the market. And if you want to be active in the US, you have HIPAA to follow reg. your data and generally need to be compliant with FDA regulations. Most other regions or countries have some kind of similar setup.</p><p>Besides the regulatory side, you are often also not dealing with a direct to consumer market where the consumer also pays for your product. Healthcare is often involving many more stakeholders, such as doctors, clinics, health insurances, and authorities. That means that the user of your product may not be the person or organization that pays for your product. And it often needs to be embedded in clinical or other workflows or be compliant with very specific regulations for people to use your product and to get reimbursed for it.</p><h3>The last years in digital health</h3><p>The last years have shown that there’s plenty of interesting topics and fields that can be improved with digital products. Lots of companies got founded and offer a plenitude of digital health apps and products. Early examples are <a href="http://klara.com/">Klara</a> (previously Goderma), <a href="https://www.mimi.io/">Mimi</a>, <a href="https://mysugr.com/de/">mySugr</a>, <a href="https://helloclue.com/">Clue</a> and others. Organizations such as <a href="https://bundesverbandinternetmedizin.de/">Bundesverband Internetmedizin</a> were very helpful for guiding those companies through the complex regulatory field.</p><p>So far companies had to be really creative to actually build a business around their products. This definitely applies to Germany with its healthcare system that is very much centered around health insurances that usually pay for most of healthcare. In other systems with a more private part of healthcare spending it’s a bit easier to build a business. Thus some companies re-focussed their business from Europe to the US. Companies that are still active in Germany and work with public health insurances usually have some kind of marketing based contract, a contract centered around a prevention case, or a special agreement with individual health insurances (in German Selektivvertrag).</p><h3>Potential of digital health</h3><p>There’s lots of regulatory hindrances for a reimbursement via the regular healthcare system. Though digital health makes it possible to have a much stronger interaction and involvement of the patient or user. With that prevention, screening, and aftercare can be done much more easily. Also treatments that take longer or treatment of chronic diseases can be done in totally different ways. The goal should be to have digital health products in parallel to all normal kinds of health products, also in Germany. Fingers crossed that the activities of the current Federal Minister of Health in Germany, Jens Spahn, pay off and introduce a more predictable and reliable pathway into the healthcare system.</p><blockquote>You use SaaS, PaaS, IaaS or DBaaS because running servers is not the core of your business. Neither is regulatory compliance. Consider Regulatories as a Service (RaaS) for your company.</blockquote><h3>What should you focus on?</h3><p>If you’re building a digital health product, you need to focus on your product and how to establish it in the market to get reimbursed for it. The regulatory side with medical product certification and data protection is important, but shouldn’t be your biggest worry. You probably don’t have physical servers in your own office and (at least in the beginning) didn’t do the whole accounting in-house, especially the annual financial statements (in German Jahresabschluss) of your company. So why should you do all regulatory, especially with the new Medical Device Regulation that’s coming next year, work in-house? You should focus on the core questions of your business. Please feel free to reach out how I can help you with your quality management system, technical documentation, or other (technical) questions. Looking forward to hearing from you.</p><p>Photo by <a href="https://unsplash.com/photos/0-SGyQFiDRI?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">rawpixel</a> on <a href="https://unsplash.com/search/photos/%22medical-app%22?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8efc2d071ae4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Machine learning for medical imaging analysis]]></title>
            <link>https://pascalwerner.medium.com/machine-learning-for-medical-imaging-analysis-faf89dd8dfa8?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/faf89dd8dfa8</guid>
            <category><![CDATA[medical]]></category>
            <category><![CDATA[automation]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[healthcare]]></category>
            <category><![CDATA[deep-learning]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Thu, 07 Feb 2019 15:38:32 GMT</pubDate>
            <atom:updated>2019-02-07T15:38:32.443Z</atom:updated>
            <content:encoded><![CDATA[<p>Everyone talks about machine learning. Why is it not already commonly used for medical imaging analysis? What are the challenges and blockers?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3a7Cx-NeYbIep2amS3SodA.jpeg" /><figcaption>Not every x-ray is as simple as this one.</figcaption></figure><p>Computers are already <a href="https://deepmind.com/blog/alphazero-shedding-new-light-grand-games-chess-shogi-and-go/">great</a> at chess, Go, and even some complex <a href="https://www.newscientist.com/article/2191910-deepmind-ai-thrashes-human-professionals-at-video-game-starcraft-ii/">computer games</a>. Especially Deepmind, a subsidiary of Alphabet is quite advanced. Also it can nicely detect whether there’s cats in videos.</p><p>Algorithms, processing power, and data are the ingredients for machine learning. One exception are situations with very clear rules like chess. Deepmind’s AlphaZero for example was training solely based on the rules of chess. Though for more complex, real-world problems that usually doesn’t apply, which means that you need data for training of your machine learning system.</p><h3><strong>What is machine learning?</strong></h3><p>From an application perspective, machine learning is split in two steps. The first is the training. And the second is the classification. The data that you usually need for machine learning are needed for training the system and validating it. The trained machine learning system is then fed with new data and gives you its output based on the training data. That means that a machine learning system can only be as good as its training data. And that also explains why machine learning systems can be racist. If you only use images of white middle-aged men to classify humans, then the system will most likely only classify white middle-aged men as humans. It’s not the machine learning system that’s racially biased, but the person building the set of training data.</p><h4><strong>What data do I need?</strong></h4><p>Regarding the kind of training data, there’s one big question. Are they labeled or not? Labeling means tagging each image with the respective classification of that particular image. A simple example would be using two different classes for classifying images. The one label is “cat”, the other label is “no cat”. And each image either has a cat or no cat.</p><p>Unsupervised learning is not as straightforward to apply as there’s no guidance for the system what it should detect.</p><p>Supervised systems are the much more systems. But as you can already imagine, besides having good data, you also need good labels. If a lot of the input data is wrongly classified, then the machine learning system can’t be trained properly.</p><h4><strong>Traditional machine learning vs new and fancy deep learning</strong></h4><p>Traditional machine learning systems consist of two stages. The first stage is about transforming the input data into simplified data that the machine learning system can work with. This is usually called feature extraction. If you have images of sticks and you want to classify it by the material, some of the features that you define could be the color, the length, or the thickness. You need to define anything that you assume will help the actual machine learning system to make sense of the input. And those features, in this case the color, length, and thickness, is being used as input for the machine learning system (e.g. a neural net).</p><p>And then, as described above, the system will use those features together with the labels, e.g. wood and metal in this case, for each individual image and get trained based on it. If you then have new images of sticks you can feed it into it. First it calculates the features and then classifies it with the trained machine learning system.</p><p>That’s at least how machine learning worked for a long time. Nowadays you hear a lot about deep learning. And the great aspect about them is that you don’t need this feature extraction step. It’s done within the deep learning system itself and doesn’t need human guidance. That’s obviously amazing because it’s not limited what humans consider useful for the classification, but can leverage any information for which the computer with all its processing power finds any correlation to the label information.</p><p>So, what’s the downside of deep learning? You need a lot of data for the training, much more than for traditional machine learning approaches.</p><blockquote>Long story short, you want a lot of high-quality labeled data to be able to use deep learning, which would make you benefit most from machine learning.</blockquote><h3><strong>Why should we use machine learning in medical imaging analysis at all?</strong></h3><p>Let’s look at how human doctors learn how to analyse medical images. Quite simplified, they look at a lot of images and learn how to classify them and what to look out for. Well, that’s not so different from how a machine learning system works. Though a computer can process many more data than a human can. A machine learning system has quite some advantages:</p><ol><li>can use all existing training data from all over the world</li><li>can be updated regularly</li><li>doesn’t get tired</li><li>it can get replicated easily</li><li>a computer can process so many more data in parallel (e.g. x-ray, MRI, US, PET)</li></ol><p>Even if doctors would never get tired, they still can only digest a specific amount of data for learning. And what would you prefer as a patient, a doctor that was trained on a subset of medical images that were available to him or a system that was trained with all medical images that are available worldwide? Also maybe you have a rare disease that your specific doctor may not have seen before. He will most likely not recognize it.</p><p>And combining several data sources, maybe for a full diagnosis a combination of your PET and MRI scan are necessary. Combining multiple, potentially multi-dimensional sources of information is pretty hard for humans.</p><p>Human doctors are still very much needed and ideally work together with a tool as described above. Human doctors can verify the results of the computer and are also much better in human communication.</p><h3><strong>Again, why is this not done yet?</strong></h3><p>As discussed a lot of data are necessary for this. And there are not many databases of medical data or images available. This is one of the reasons for Deepmind <a href="https://www.digitalhealth.net/2017/03/deepmind-mustafa-suleyman-interview/">working the the British NHS</a>, as they are a nationwide health system that can give access to huge amounts of data.</p><p>There’s a lot of great researchers out there, also processing power is not the limit. The main bottleneck to make progress on the topic of machine based analysis of medical images is to have sufficient data available.</p><p>Photo by <a href="https://unsplash.com/photos/ouyjDk-KdfY?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Harlie Raethel</a> on <a href="https://unsplash.com/search/photos/xray?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=faf89dd8dfa8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Current state of GDPR, part 2: how to be reasonably compliant?]]></title>
            <link>https://pascalwerner.medium.com/current-state-of-gdpr-part-2-how-to-be-reasonably-compliant-e3f55835a9d0?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/e3f55835a9d0</guid>
            <category><![CDATA[data-protection]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[gdpr]]></category>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[user-experience]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Sun, 13 Jan 2019 19:41:58 GMT</pubDate>
            <atom:updated>2019-01-13T19:43:20.230Z</atom:updated>
            <content:encoded><![CDATA[<p><strong>Current state of GDPR, part 2: how to be reasonably compliant?</strong></p><p>This is the second part of a series on the current state of GDPR and discusses a risk-based compliance approach. The <a href="https://medium.com/@pascalwerner/current-state-of-gdpr-two-sources-of-fear-7cc77c48f1a">previous and first part</a> was about the two different fears that are caused by GDPR which can even stifle innovation. The third and last part will be about some ideas how a long-term vision of data protection can look like.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*raKKXM9sfZtPDrlqJH7yEA.jpeg" /></figure><p><strong>Again, what’s the issue about GDPR?</strong></p><p>GDPR itself and its aspiration to improve data protection and users’ rights on their personal data is a good thing. From a legal perspective the main issue would be if you’re not compliant. Due to still missing significant fines from the supervisory authorities, quite some companies are waiting to see whether they should invest into GDPR compliance at all. But that’s mainly an enforcement issue and can be changed quite easily.</p><blockquote>“There’s no prize for over-compliance, but an increase in implementation and recurring costs”</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5wwBHnoo0f1onEw7xtn3sw.png" /></figure><p>The bigger issue about GDPR is missing best practises and court rulings which would give interpretation guidance. Looking at the small illustration above, you generally want to be in the green zone, to be compliant at reasonable costs. Though that’s pretty hard because most advise that you get is coming from lawyers and consultants. Lawyers often interpret law from a worst-case scenario perspective and, given no existing court rulings, are interpreting GDPR in the strictest way. And consultants make money by having more work. So I’d not be surprised if there’s regularly a tendency to the yellow zone.</p><p><strong>Risk-based GDPR compliance</strong></p><p>How does this green zone look like? In theory it’s quite easy. You need to understand the regulations in detail as well as know your product and business situation. You want to find the sweet spot in between your legal, product, and business requirements.</p><p>Three guiding questions that usually help a lot are:</p><ol><li>What is a good user experience? Neither do you want your users to be negatively surprised by seeing unexpected ads nor do you want to annoy them with huge consent forms.</li><li>Is it clearly non-compliant? There are a few topics in the GDPR that you just need to do.</li><li>Are you compromising on IT security? Don’t do that. You don’t want to be responsible for the next leak of personal data.</li></ol><p><strong>Implementation of risk-based compliance</strong></p><p>Having respect for the GDPR, but no fear sounds good, right? Following that approach you’ll most likely anticipate best practices before they materialize. You can regard yourself as innovative and you don’t invest more resources than necessary. And by turning fear into respect you can enable your employees to do their work and take good decisions in an increasingly complex world.</p><p>So why don’t more companies do that? One reason is surely that any change is hard, especially the bigger the company. But another reason is most likely lack of knowledge about doing reasonable GDPR compliance. And especially the latter can be changed. But any progress in this regard can’t happen without the backing of a company’s (top-) management.</p><p>Credit for the Photo by <a href="https://unsplash.com/photos/U5y077qrMdI?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Mark Duffel</a> on <a href="https://unsplash.com/search/photos/rules?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e3f55835a9d0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Current state of GDPR — two sources of fear]]></title>
            <link>https://pascalwerner.medium.com/current-state-of-gdpr-two-sources-of-fear-7cc77c48f1a?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/7cc77c48f1a</guid>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[gdpr]]></category>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[fear]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Sun, 13 Jan 2019 13:29:37 GMT</pubDate>
            <atom:updated>2019-01-21T19:42:47.524Z</atom:updated>
            <content:encoded><![CDATA[<h3>Current state of GDPR, part 1: two sources of fear</h3><p>This article is split in three parts. The first part will discuss the current state of GDPR. <a href="https://medium.com/@pascalwerner/current-state-of-gdpr-part-2-how-to-be-reasonably-compliant-e3f55835a9d0">The second part</a> is about reasonable compliance and a risk-based approach for that. The third part collects some ideas how a long-term vision of data privacy can look like.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*cnWe9CM6F1-quYSc43-1Tw.jpeg" /><figcaption>Besides the obvious fear of supervisory authorities, there’s the more hidden fear of compliance programs in regards to GDPR, especially the latter can be detrimental for innovation.</figcaption></figure><h3>Short overview</h3><p>Everyone should have heard of the General Data Protection Regulation (GDPR) by now and surprised by the strength of its hype in mid 2018 and its non-existence in public discussions now. GDPR can be roughly summarized by the pre-existing data protection regulations combined with draconic fines. And it applies to companies world wide if they target European customers.</p><h3>Why was GDPR a big topic</h3><p>Exactly those potential fines combined with a lack of clear guidance led to a surprisingly huge wave of public attention. But then on May 25, 2018 GDPR became active and people realized that the world is not going to stop. Supervisory authorities were said to start enforcing GDPR sometime later in 2018. With the immediate danger disappearing, the public attention declined rapidly. Though there’s still a strong source of fear around that even stifles innovation. To explain that we first need to look deeper at some basics.</p><h3>What is over-compliance?</h3><p>Let’s first look at different levels of GDPR compliance. For simplification, let’s assume that compliance is a one dimensional axis. On the one side you have clear non-compliance. The further you go to the right side, the higher the compliance. Though there’s neither a price for the best GDPR compliance nor an absolute definition of it. It depends a lot on your business and technical setup. Everything that’s not necessary and that you still do could be called over-compliance. And that causes additional costs and can easily be used as a counter-argument on new products. Next we’ll have a look at the sources for fear.</p><h3>Two sources of fear</h3><p>The obvious fear from GDPR was induced by the high potential fines from the supervisory authorities. Though that source is pretty empty right now, at least until first bigger fines were imposed.</p><p>The other source of fear is of private or internal nature. It’s based on contracts between companies or internal compliance programs. Those contracts could be between your startup and an investor or bigger customers. Internal compliance programs are usually rather existing in bigger corporations. You personally may have not a lot to gain from a specific new potential product that you could pursue, but a lot to lose if you don’t comply with company policies. That’s why GDPR is a pretty big topic in larger companies despite the lack of public presence.</p><h3>Outlook and what to do as a company</h3><p>So why is this the case? One big reason is the lack of best practices. So far it’s mainly lawyers that tell you what you need to do to be GDPR compliant. Though the job of a lawyer is to explain you what to do for full compliance, they can’t advise you on a middle-path. But exactly this middle-path is necessary to still have room for your company to prosper.</p><p>Don’t get me wrong. I’m a big fan of strong IT security. It’s quite annoying to regularly read about data leaks of gigantic datasets. <a href="https://medium.com/@pascalwerner/current-state-of-gdpr-part-2-how-to-be-reasonably-compliant-e3f55835a9d0">The second part</a> of this series will describe how such an approach of reasonable compliance can look like.</p><p>Also, most privacy experts are still expecting more substantial fines to be imposed in the first half of 2019, setting a precedence for the respective industry. So stay tuned.</p><p><em>Update February 21, 2019<br></em>The French supervisory authority <a href="https://www.cnil.fr/en/cnils-restricted-committee-imposes-financial-penalty-50-million-euros-against-google-llc">CNIL just imposed a fine</a> of 50 mio € on Google for its noncompliant Android user on-boarding flow. And several German supervisory authorities <a href="https://www.handelsblatt.com/politik/deutschland/datenschutzgrundverordnung-behoerden-verhaengen-erste-bussgelder-wegen-verstoessen-gegen-dsgvo/23872806.html">imposed about 41 fines</a> on German companies. So that article’s conclusion can be basically rewritten.</p><p>Credit for the Photo by <a href="https://unsplash.com/photos/MTjP7QxJESM?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Tyler B</a> on <a href="https://unsplash.com/search/photos/dangerous?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7cc77c48f1a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Performance marketing vs. data privacy (GDPR)]]></title>
            <link>https://pascalwerner.medium.com/performance-marketing-vs-data-privacy-gdpr-4b3cccfad278?source=rss-e144ce40a263------2</link>
            <guid isPermaLink="false">https://medium.com/p/4b3cccfad278</guid>
            <category><![CDATA[marketing]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[adtech]]></category>
            <category><![CDATA[compliance]]></category>
            <category><![CDATA[gdpr]]></category>
            <dc:creator><![CDATA[Pascal Werner]]></dc:creator>
            <pubDate>Tue, 20 Nov 2018 19:05:32 GMT</pubDate>
            <atom:updated>2018-11-20T19:19:45.324Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XQ-Jm8pc0HAsF_hJw0eFhA.jpeg" /></figure><p>There’s a lot of talk and articles on digital marketing and how to comply with GDPR, the recent European data privacy regulation. But there’s not so much talk about long-term implications and possible developments for marketing efforts. To be clear, if you’re looking for specific advice on how to make your current marketing efforts GDPR compliant, this article is not for you.</p><p>Long-term, GDPR should have triggered a development towards less scary marketing methods. Though for this to really solidify it will take some more time. So far GDPR clearly favours big players, namely Facebook and Google, simple due to their immense financial resources. Smaller ad-tech companies had to cease business because they didn’t manage to become GDPR compliant. Facebook and Google have the funds to have reached compliance, though it may be a rather superficial compliance. Superficial compliance refers to making an inherently non-compliant business model seemingly data privacy compliant, without following the actual intent of the regulation. While this is an issue in almost all existing industries, high-margin business models in the region of 100 billion USD are obviously most resistant to change.</p><p>The whole ad-based business model of Facebook and Google seem to be build on non-transparency. With GDPR in place, non-transparency translates into non-compliance. GDPR requires certain user rights, which also includes the possibility to revoke consent. With something that intrusive as personalized ads based on data from various data sources, explicit consent via an opt-in would be required to be compliant. Facebook and Google are obviously not interested in that.</p><p>Over time, data privacy friendly alternatives will get established. This may happen due to enforcement by regulatory authorities or simply by user’s choice. Big tech are obviously on the radar of data protection authorities. While the ICO, the UK’s data protection authority, <a href="https://ico.org.uk/action-weve-taken/enforcement/aggregate-iq-data-services-ltd/">has issued the first formal enforcement action under GDPR</a>, actions from other European data protection authorities are expected early 2019.</p><p>The big question is: what comes next? Companies are looking for places where they can spend their marketing budgets. Also people don’t want to see random and totally unrelated ads. <a href="https://hbr.org/2018/05/how-gdpr-will-transform-digital-marketing">An article on HBR</a> talks about contextual ads, where the ad content is not based on the person, but on the environment he’s getting the ad served. This is already done by some companies, but will certainly attract more attention.</p><p>What are your thoughts on that? Please post below or reach out to me!</p><p>Photo by <a href="https://unsplash.com/photos/u5Zt-HoocrM?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Dayne Topkin</a> on <a href="https://unsplash.com/search/photos/data-privacy?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4b3cccfad278" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>