Category Archives: Healthcare Industry

Social Media Analytics for healthcare – Human involvement in the 4-I continuum

Analytics, as known in the digital data parlance, is perceived to be associated with or rather dependent on tools and technologies. Undoubtedly, various software tools and technologies have significant roles to play in analytics but their contribution is limited to data extraction, organization, colleting, counting and to a limited extent, recognizing textual patterns.

Current technologies cannot replace humans in executing critical aspects of analytics; when it comes to managing ambiguity and complexity, human intelligence is indispensible. For instance, it is difficult for any software to define the problem, interpret results in the context of the problem, filter the relevant information from large chunks of data and derive actionable insights to mitigate business risks or take corrective action.

As compared to consumer goods, monitoring and analyzing social media information in healthcare and pharmaceutical industry is far more complex due to various reasons such as regulatory constraints, privacy and confidentiality issues, non-standard patient experiences, complexity in understanding symptoms, side effects, reaction to drugs, reasons for patients switching brands etc.

Broadly, social media analytics process involves converting information into intelligence, intelligence into insights and identifying initiatives based on these insights. As we traverse through the 4-I continuum of information-intelligence-insights-initiatives, it is critically important to understand the roles of human analysis.

The art lies in understanding what can be achieved using software and what should be done by humans. Here’s my take on that.

4-I Continuum of Social Media Analytics

4-I Continuum of Social Media Analytics

Before we start, I would like to mention that it is critically important to define the business problem and set the direction of research; needless to say, both have to be done by humans.

Information stage: Although low as compared to the advanced stages, the human involvement is needed in defining the objectives of the research, creating keyword taxonomy and feeding it correctly in the system using Boolean search logic. Several listening-software products such as Radian6 claim to provide real time social media data analytics, automatically and conveniently. In my experience, automated listening software do a fair job at discovering data, are good at extracting data but not so good at giving out meaningful analytics. They certainly do the counting part well. What the software gives you depends on what you ask. It is important that the right instructions are received by the software. Importance of defining the taxonomy that covers all aspects of the problem cannot be overemphasized.

Intelligence stage: Use of the term “intelligence” should be sufficiently indicative of the strong human involvement is required at this stage. The extracted data includes important information, some irrelevant data and some junk. Various methods are used to filter out the junk as well as separate the irrelevant material. Let me elaborate on “relevance” here; information may be relevant to the keywords, but not necessarily relevant to the objectives. For instance, if “skin cancer” is a keyword, then information around skin disorders (not necessarily cancer), is relevant to the keyword “skin” but not relevant to skin-cancer. Further, data around skin care products or animal-skin products would qualify as junk. Such distinctions can only be humanly made. Boolean search will allow data filtering based on such criteria, but high accuracy is rare.

Insights stage: Insight generation is interpreting the information in the context of the defined problem. At this advanced stage of analytics, human involvement becomes even stronger. At this stage, the relevant information is categorized, tagged, analyzed and consolidated. Combining insights helps solve the problem. At this stage, technology can play a diminished but nevertheless significant role by offering insights at one place. For example, a consolidated mash-up of charts in a dashboard can show sales trends in one chart, social media sentiment trend in another, a tag-cloud of side effects in a third and switch-over patterns in a fourth. All these put together may help in understanding the cause of the problem.

Initiative stage: This is a purely human activity since it involves taking decisions around the “now what” part i.e. action. The information has been extracted, intelligence gathered, insights derived, problems identified and now it is time for initiatives i.e. finding and implementing a solution to the problem.

For instance, if there is substantial noise by a social advocate around side effects of a particular drug on Twitter, the brand team can implement digital initiatives and engage the advocate for spreading information around managing side effects using (hypothetical example) a vegan diet.

In summary, for efficiently implementing social media analytics programs to solve business problems, managers have to optimize the use of technological and human resources in the process, as they traverse the continuum of information-intelligence-insights-initiatives.

10 lessons I learned in healthcare social media analytics

It is said that experience is the best teacher. Here, I want to share the 10 best lessons I learned in the past 1.5 years, while establishing Social Media Monitoring and Analysis practice serving the global pharmaceutial and healthcare industry.

  1. Machines rarely deal well with human emotions. Natural  language processing  software or textual analysis software may be good at recognizing  patterns of how certain keywords appear in certain sequence and how that can be interpreted, but when it comes to understanding opinions of patients or reasons for switching drugs or therapies, they rarely do a good job. That is why you need human analysts or medical experts  in the team who exactly understand the content and tell the truth.
  2. The problem isn’t always “big data”. Most approaches to social media analytics start with the assumption that data volume is very high and users are increasingly conversing, thus compounding the problem. This may be true of certain consumer goods, but when it comes to healthcare and prescription drugs, big data is not always the issue. In fact, for prescription drugs, data volume doesn’t become “big data” during pre-launch, launch and post launch phases; it takes years before the drug is adopted by a mass of people and data becomes big. By then, there is not much left to change.
  3. Statistical sampling doesn’t work in qualitative analytics. It is always safer to look at the entire quantum of data for drawing conclusions; a sample section of the data rarely gives an accurate big picture. This is possible to be done during the decision making phase, since the data is not yet big. If you want to be confident about the accuracy and quality of analysis, there is substitute to manual analysis of large part of the data. Afterall, we are talking healthcare here; the analysis may have serious consequences.
  4. Standard syndicated reports are limited in their value. Syndicated brand reputation report covering entire competitive landscape are good to have, but they rarely offer insights. They don’t solve problems or provide competitive advantage. Deeper dives are needed to reveal brand specific insights and most clients are often willing to pay for custom investigations.
  5. Influence of patient generated content extend beyond geographic limits. Clients seem to believe that social influence is local. So most affiliate (country level) offices aren’t interested in what is being said internationally, although the brand is global. The reality is, patients read everything that appears in searches and don’t necessarily spend time in selecting comments only from their geography. In order to understand what influences patients in a certain geography or country, it is better to focus on the content being consumed there, instead of focusing on the content being produced there.
  6. Social media analytics is not an event; it is a process. Short term analysis of patient comments done within a time window, may reveal insights. But that may be only a section of truth. In order to understand long term trends, changing opinions, shifts in brand perception and conversation triggers that caused them, continuous monitoring is needed.
  7. Adverse events reporting using social media is rare. Patients rarely use social media for actively reporting adverse events. Even if we assume that they inadvertently do so, most of the times, social media content doesn’t qualify as reportable adverse event.
  8. Even if there are some AEs, reporting can be easily managed. Some may not buy point 5 and may believe that there may be some patients who report AEs using social media. There is so much to be gained from listening to consumers that there’s no point in sacrificing it all for concerns around AE reporting. Social media AEs are easy to manage. Define a process. Create protocols. Delegate responsibilities and put a team together. Treat social media as another channel for reporting AEs, just like other channels such as a reporting form on the website, or a call center number or through physicians.
  9. It is critical to think from the consumer’s perspective. Social media is not about large corporations and drug manufacturers; it is about people and their lives. Patients talk more about their own condition and their experience with your brand. Social media is deeply integrated into the patient’s treatment journey. Patients use social media as soon as symptoms start appearing and continue right through, until they are cured or come to manage a lifelong condition . This way of thinking should lead to social media initiatives targeted at patients at each stage of the treatment journey.
  10. Integrated marketing is the right approach. To build a sustainable competitive advantage using social media, we have to think holistically. Social media is a part of the digital marketing strategy, which in turn is a part of the larger brand marketing strategy. You don’t have to do all the marketing in-house. Partner with an agency which can work with you across the digital/social media marketing continuum i.e. social media monitoring and analysis, integrating insights into the marketing mix, social media marketing and engagement and dollar impact measurement.

(Views are personal.)