Microlearning: From question to insight

Step 1: Frame the question

Before jumping into the data, it’s essential to pause and ask: What exactly is the question we’re trying to answer? Framing the right question not only guides your analysis but also ensures that the data you collect and how you interpret it are meaningful and actionable. It's also important to consider and be realistic about the problem you're trying to solve.

From here, we can drill down by framing the question: What exactly do we mean by wait time? Which age groups should we report on? How do we define arrival and seen?

These definitions matter because they influence the data points you use and the conclusions you draw. Clear questions make clear data needs.

---

Step 2: Find and make sense of your data

You start investigating

You pull the ED system logs and find raw records. You’re looking at patient arrival times — that’s the data. But how do you know what the timestamp means? Is it the arrival at triage? At registration? What’s the time zone? Is it formatted as AEST time?

Patient ID	DOB	Arrival Time	Seen Time	Status
1001	1991-09-08	2025-06-01 18:46	2025-06-01 19:47	Seen
1002	1954-12-02	2025-06-01 15:16	—	Left AMA
1003	NA	2025-06-01 08:06	2025-06-01 10:01	Seen
1004	2000-07-23	2025-06-01 09:13	2025-06-01 09:39	Seen
1005	1960-08-05	NA	2025-06-01 16:25	Seen
1006	2014-03-12	2025-06-01 08:03	2025-06-01 09:07	Seen
1007	2014-08-22	2025-06-01 09:55	2025-06-01 11:34	Seen
1008	1963-11-03	2025-06-01 14:35	2025-06-01 15:07	Left AMA
NA	1942-01-03	2025-06-01 16:18	2025-06-01 16:39	Seen
1010	1961-05-15	2025-06-01 15:49	01-06-2025 16:56	Seen

That’s where metadata comes in. Metadata is data about data — information that describes and gives context to the data. This enables proper understanding and interpretation of the recorded data, and guides data usage and analysis.

Data	Metadata
Data values include: patient vital signs (such as heart rate, blood pressure, respiratory rate), laboratory test results (such as blood glucose levels, cholesterol levels), medication orders (such as dosages, frequency, and drug types), and clinical observations (such as wound descriptions or radiology findings).	Metadata offers vital: contextual details, definitions, standards, and formats.

Let’s see how this works in practice. Below is a data point — can you interpret it without knowing the context? Without metadata, this number is incredibly hard to interpret.

Try to guess what this figure represents. Could it be related to Patient ID, DOB, timestamp or something else entirely? Select the arrow to view the metadata attached. Then, answer some questions about data and metadata for our scenario.

---

Step 3: Identify the data quality & get the right structure

Data appears in many different forms and formats. Let’s break down the types of data you might find. Healthcare data can be broadly classified into two types: structured and unstructured forms. Select the cards for information on each data type.

All the data for our scenario is structured. However, for some practice, let's examine a range of other data sources and determine whether you would consider them to be unstructured or structured data. Select your answer below and the select the arrow to move to the next question:

Structured data is easier to analyse and merge (e.g. summing lab results), but unstructured notes often contain rich detail (e.g. social history, symptom descriptions). In practice, most EMRs mix both: vitals, test results and coded problems are structured, whereas clinician observations and patient stories are unstructured. This distinction matters because unstructured data requires natural language processing (NLP) or manual review to extract information, adding time and cost.

Even though structured data is easier to analyse, it's often entered incorrectly or is incomplete. Based on obtaining these data points, can we calculate the average wait time now?

---

Step 4: Clean & prepare the data

Now that you’ve gathered the data and understand its structure, it’s time to transform it into something useful. Raw data often contains gaps, inconsistencies, or irrelevant records, and jumping into analysis without addressing these can lead to misleading conclusions.

Time for a spring clean

Here’s what you do:

1. Calculate age based on the DOB
2. Calculate wait time (Seen Time – Arrival Time)
3. Group patients by age (e.g., 0–18, 19–39, 40–59, 60+)
4. Exclude patients who left without being seen or who have incomplete data

Now your dataset is cleaner, more structured, and ready for insight generation. 

Patient ID	Age	Age Group	Wait Time (mins)	Status
1001	60	41–60	79	Seen
1002	22	21–40	—	Left AMA
1003	45	41–60	40	Seen
1004	75	61–80	75	Seen
1005	30	21–40	30	Seen

---

Step 5: Analyse the data

With your clean dataset in hand, it’s time to explore: What story is the data telling you? Start by summarising key metrics.

What’s the insight?

Older patients tend to wait longer.

Now we can explore triage practices, staffing patterns, or seating availability. With further analysis of the data, you can derive that:

• Peak wait times are between 10 AM and 2 PM.
• Staffing doesn’t match peak demand.

---

Step 6: Gain insights (decisions, decisions)

Now that you’ve identified patterns in the data, the real work begins: deciding what to do about it. Analysis is only powerful if it leads to meaningful decisions. You have several choices. Select each of the options below to learn more. To return to the list of options, select the home screen.

---

Key takeaway

Data analysis is only the beginning. Real impact occurs when insights shape your decisions, and when those decisions are measured and improved over time.

Want to dive deeper into data analytics and explore how to implement AI solutions in healthcare? Read more about the full course here: Transforming Healthcare with Data Analytics and AI, or sign up now.