The Four Types of Health Data Analytics

Health data analytics is not one activity. A useful way to organise it is by the question being asked. The four types are descriptive, diagnostic, predictive, and prescriptive. They are a sequence of questions, not four reasons to jump to a conclusion.

Descriptive analytics: what happened?

Descriptive analytics summarises what has already been recorded. It answers questions such as how many, how often, and how the figure changed.

A fictional primary healthcare centre records confirmed malaria cases among outpatients:

The description is straightforward. Cases were fairly steady in January and February, then rose in March and rose again in April. The four-month total is 143 cases. That is what the register shows. It is not yet an explanation.

Diagnostic analytics: why did it happen?

Diagnostic analytics looks for reasons. In this clinic, staff might ask whether the rise followed the start of the rains, a stockout of insecticide-treated nets earlier in the year, a change in testing, or more people attending because a outreach day was held.

These are possible explanations. They are not findings until they are checked against other records: the rainfall period, the store ledger, the laboratory tally, and the attendance register. More cases can also mean that testing improved, not that malaria transmission increased.

Predictive analytics: what is likely to happen?

Predictive analytics uses patterns to say what may happen next. If this clinic has seen malaria rise in March and April for several years, a planner might expect May to remain high and might forecast the number of malaria tests and artemisinin-based combination therapies needed.

A forecast is a reasoned expectation. It is not a fact about the future. It should say what information it used and what could make it wrong, such as a change in rainfall, a stock delivery, or a change in how cases are counted.

Prescriptive analytics: what should we do?

Prescriptive analytics turns a checked finding into an action. If the rise is supported by the register, and the store shows that malaria medicines will not cover the usual May demand, the action might be to request more medicines before the stock runs out, and to confirm that nets are available for the households the programme is meant to reach.

The action should name who will do it, by when, and how the clinic will know whether it helped. Prepositioning medicines is a supply action. It does not, by itself, prove that the quality of clinical care has changed.

Facility case

The officer-in-charge sees the April figure of 60 and says, “The rains caused this. Order double the medicines and report that malaria control has failed.”

A more careful reading separates the four questions. What happened: recorded cases rose from 18 in February to 60 in April. Why: not yet known. What may happen: if the usual season continues, May may also be busy. What to do now: confirm the count, check the medicine ledger, and place an order that matches the verified need.