6 skills/traits health data analysts need

In a previous post, I wrote about the 5 reasons to switch to a health analytics career.

To pursue a career in health analytics, some quantitative skills are necessary. However, to really thrive in a health analytics role, I believe softer skills and personal traits are also important. Here are 6 skills/traits health analysts need to succeed:

1. Curiosity

Curiosity turns boredom into joy, mundane into germane, mediocrity to super stardom. If you're curious about something, you won't mind spending time learning and figuring out solutions. Healthcare is complex, intersecting medicine/science, politics, technology and finance. This complexity coupled with the possibility to truly impact patient's lives make me endlessly curious about healthcare.

These days, it's nearly impossible to read the news without seeing some talk about health reform, about chronic diseases, about the opioid pendemic etc. Behind each of those stories, is a whole lot of policy and data analysis. You could be the analytic detective behind those stories. Everyone has some personal/familial experience with illness or death. Trust me, when you know your work saves lives, you cannot wait to get up in the morning to get to work!

2. Knowledge of the dynamics in health care

There are many stakeholders in healthcare (future post to follow), each with their goals, expertise, incentives, attitudes. When people talk about actionable analysis, they are talking about analyses that are acutely aware of these dynamics. If you're blind to these driving forces behind decision making, you will miss the WHY of the analysis was needed and how your results will be perceived.

I found that my masters degree in health policy gave me some tools but it's really the discussions with colleagues and reading of (mostly free) published high quality papers that helped me understand the nuances in healthcare.

3. Numeracy

You have to numerate to be a data analyst. That does not mean you don't need a calculator when it comes to paying bills at a restaurant. But it does mean you should be very comfortable thinking abstractly through average, trends, drives of differences. Experience analysts develop a sense of whether a set of results are right, and an innate sense of the most practical, fastest way to perform the analysis. Here is a post on feature engineering and one case study on how to use analytics to manage a medical practice.

You will learn the most on the job. While I have some economics and statistics training, truth be told, the level of maths you will do most of the time is basic, so any maths, computer science, epidemiology, finance training should prepare you adequately. Here is a short list of key statistical concepts you should know.

4. Knowledge of health data

Health data has a lot of distinct features compared to data from other industries. Some data types are for billing, while others for clinical information, which can be highly nuanced. Understanding how data was generated, how it was processed can also give you a sense of how much you should trust it and how to use it.

Data quality is a major issue in healthcare. On average, a project probably consists of 20-30% of time dedicated to exploring, cleaning, ingesting data. Having a healthy dose of awareness of these potential issues upfront could save you lots of headache.

From my experience, this is one area that leaving to learning on the job will likely be very slow and often incomplete. I've been doing this for almost 15 years and I'm still learning constantly about different aspects of healthcare data everyday.

I will be dedicating a series of posts on this topic in future. See the sources of healthcare data and types of healthcare data, such as diagnoses, procedures, drugs, laboratory codes, etc.

5. Interpersonal skills

You will likely work with people who are not very analytically inclined, such as doctors or pharmaceutical reps. You will likely work with people with lots of opinion and often huge egos, such as patient advocates and surgeons. You could also quite likely step on many toes as your analysis shines light in places that like to be kept in the dark...

All that makes for an interesting and meaningful challenge. Your interpersonal skills of being able to step into other people's shoes, explain concepts/results in ways they understand and appreciate, seek common grounds/trust and be able to convince them of your recommendation would set you apart from a backroom number cruncher... Here are some practical considerations that will help you in health analytics.

6. Software

I place this last because I believe if you're curious about healthcare, you will figure out software requirements and learn those as needed. There are plenty open source software and online learning platforms, so access/cost is no longer an excuse.

The software tools I use are MS Excel (60%), SQL (25%), Machine learning/data visualization/other (15%). Most of the healthcare data I deal with are structured data, such demographic/membership, claims, EHR and financial data. The mixture of software used satisfies my needs right now, of being able to pull semi-summarized data from large databases using SQL, to explore, clean, analyze the data in MS Excel and then to use ML or data visualization tools to further enhance the analysis.

Here are some choices:

• MS Excel is the grand daddy of analytics tools. It's versatile, great for tinkering on analyses but it can be slow and very error prone. See my course on best practice analytic habits to professionalize your work. Quick tip: install the 64bit version which enables Excel to use more of the computing power of your processor.
• Data base management: Postgres SQL for database work and extraction. SQL is super easy family of languages to learn. Any health data analyst/informaticist should know SQL. You type it as you would read.
• Data visualization: Tableau for data visualization. They have a few lower cost options, academic/public/non-profit. There are other options out there, especially for dashboard building, such as IBM Cognos.
• Stats/Data mining: Knime for quick data mining, ML - it's free, scaleable and has a great GUI. R or Python are great tools for more complex statistical analyses. (SAS probably is the go to for many companies, but the cost is prohibitive, and frankly most users only use a tiny fraction of the functionalities.)

Please subscribe to receive future newsletters! Thanks for reading.