What Is Prescription Data & Where Does It Come From?

Here is the question that most Pharma analysts cannot fully answer even after years on the job: when you pull TRx data for a drug in IQVIA, what exactly are you looking at? Where did that number come from? Which pharmacies contributed to it? How old is it? And what might be missing?

Prescription data is the single most used data source in commercial pharmaceutical analytics. Market share calculations, sales force call planning, launch tracking, competitive intelligence, all of it flows from prescription data. And yet most analysts who use it daily have only a partial understanding of how it is collected, processed and delivered.

That partial understanding leads to real analytical errors: misinterpreting data lags as market shifts, over counting prescriptions because of projections methodology, under counting because a channel is missing. This post gives you the complete picture.

What a Prescription Record Actually Is:

Let us start at the very beginning. A prescription, in the data sense, is a record of a drug being dispensed to a patient at a pharmacy. It is not the act of physician writing a prescriptions, it is the moment a patient picks up ( or receives by mail) a drug.

This distinction matters. A physician can write a prescription that is never filled, the patient decides not to pick it up, cannot afford the copay, or gets denied by their insurance plan. That write event never appears in prescription data. Only filled prescriptions, actual dispensing events are counted.

Every dispense event at a US pharmacy generates a transaction record: which drug(identified by its NDC code), which patient (de-identified), which prescriber (identified by their NPI), on what date, in what quantity, for how many days of supply. This is the atomic unit of prescription data.

The 5 Step Data Origin Chain:

How does that dispensing event at a pharmacy in rural Ohio end up in an analyst’s Tableau/PowerBI dashboard in New York/New Jersey? It travels through a 5 step chain:

Step1: The physician writes the prescription

The journey begins when a healthcare provider write a prescription, digitally in most cases via electronic prescribing (e-Rx) through a platform like Superscripts, or on paper for certain controlled substances. The prescription travels to the designated pharmacy, either electronically or physically.

Step2: The pharmacy dispenses and files a claim

When the patient arrives to collect the drug, the pharmacist processes the prescription. A real-time electronic claim is submitted to the patient’s PBM for adjudication. When the claim is approved and the drug is dispensed, the pharmacy system records the transaction: drug, quantity, days supply, prescriber, date, payer information.

Step3: Data aggregators collect from pharmacies

Companies like IQVIA & Symphony Health have data purchase agreements with pharmacy chains, pharmacy benefits managers, and other dispensing entities. Under these agreements pharmacies transmit their dispensing records, typically weekly to the aggregator. IQVIA’s US pharmacy network covers approximately 90% of all retail prescriptions. The remaining 10% must be estimated through projection.

Patient identifying information is removed or encoded before transmission to comply with HIPAA privacy requirements. The prescriber NPI is retained, the prescribing physician is not protected health information which is what makes NPI level data analytically useful.

Step4: Aggregators process, project, and enrich

Raw pharmacy data is incomplete and inconsistent. The aggregator must:

1. Clean duplicates and correct obvious data errors.
2. Match prescriber NPI numbers to physician master files(adding speciality, address, and demographic information).
3. Project the data to account for the approximately 10% of pharmacies not directly captured. It is usually done using statistical models based on the pharmacies that are captured.
4. Assign geographic codes: Zipcode to territory to region.
5. Classify prescriptions as TRx, NRx, or NBRx using patient history within the dataset.
6. Apply payer classification: Commercial, Medicare, Medicaid, Cash.

The projection methodology — what you should always ask

Because no aggregator captures 100% of pharmacies, all national prescription data involves statitical projections. IQVIA projects from its captured pharmacy panel to estimate total market volume. The projection methodology is proprietary and not fully transparent. In practice, IQVIA’s projection is generally considered reliable at the national and regional levels, but can be less reliable for small geographic areas, niche speciality drugs, or drugs heavily dispensed through channels with lower capture rates (like certain speciality pharmacies). When you use IQVIA data, you are always working with a projected estimate, not a census.

Step5: Manufacturer receive & load the data

Manufacturer subcribe to IQVIA & Symphony data products, typically paying millions of dollars per year for access to national prescription data. The data is delivered as weekly or monthly files, loaded into the manufacturer data warehouse(commonly Snowflake), and made available through BI tools like Tableau or PowerBI.

The entire chain from dispensing event to analyst dashboard typically takes 2–4 weeks. This lag is one of the most important limitations of prescription data and is covered in detail below.

The 4 Prescription Metrics Every Analyst Must Know

The language of prescription data revolves around 4 key metrics. Knowing what each one measures, and what it does not, is foundational to correct analysis.

TRx — Total Prescriptions:

TRx counts all prescriptions dispensed for a drug in a given period(new patient prescriptions & refill prescription combined). It is the most commonly reported volume metric and the one most stakeholders mean when they say ‘how many prescriptions did we write last week’.

TRx is a lagging measure of commercial performance. It reflects both the success of acquiring new patients and the success of retaining existing ones. A growing TRx with flat NRx indicates that patient retention is strong but new patient acquisition has stalled (a different strategic problem than declining TRx caused by poor persistence).

NRx — New Prescriptions:

NRx counts prescriptions written for patients who have not had a prescription for this drug in the past 6 months (the specific loopback window varies by vendor & drug class). NRx is the leading indicator of new patient acquisition, it tells you whether physicians are starting new patients on the drug.

NRx is particularly important for launch analytics. In the first months after a launch, almost all TRx is NRx, there are no refills yet because no patients have been on the drug long enough. As the brand matures, refills accumulate & NRx become smaller propotion to TRx.

NBRx — New to Brand Prescriptions:

NBRx counts prescriptions for patients who have never previously received this specific brand. This is different from NRx: a patient switching from a competitor brand is an NBRx but not an NRx(they have been in the class, just not on this brand). NBRx is the primary KPI for launch tracking & competitive switching analytics.

The NBRx metric helps answer the question: where are new patients coming from? Are they treatment native patients starting therapy for the first time(a different commercial signal than patients switching from a competitor drug)?

The TRx/NRx Ratio — Persistence Signal

The ratio of total to new prescriptions over a period is a proxy for patient persistence(how long patient stay on therapy). A TRx /NRx ratio of 4.0 means that for every new prescription, there are approximately 4 refills, indicating patients average roughly 4 fills before discontinuing. A rising ratio indicates improving persistence; a falling ration indicates patients are not refilling as long as they used to.

NRx timing trap — a common analytical mistake:

NRx has a methodological quirk that trips up many analysts: a patient who discontinued therapy 7 months ago and restarts is classified as NRx (because they have not had a prescription in the past 6 months). This means NRx counts not just genuinely new patients but also returning patients who lapsed. In therapeutic areas with high discontinuation and restart rates like depression, ADHD, pain management, NRx can overstate true new patient acquisition. Always segment NRx by treatment history when this dynamic is likely to be relevant.

The Data Lag: Why Prescription Data Is Never Real Time

One of the most important limitations of prescription data and one that catches analysts off guard is the lag between a dispensing event and when that event appears in the data.

The typical lag for IQVIA weekly NPA data is 2–4 weeks. The sequence: a patient fills a prescription on Monday. The pharmacy transmits data to IQVIA the following weekend. IQVIA processes and projects the data over the next week. The file is delivered to the manufacturer mid-following week. After ETL & dashboard refresh, the analyst sees the data roughly 3–4 weeks after the original event.

This lag has important practical consequences:

1. Launch tracking: The first week of a product launch will not appear in data for 2–4 weeks. Using 852 wholesaler inventory data(which arrives faster) as an early proxy is standard practice for launch analytics.
2. LOE analysis: Prescription data will not show the full impact of a generic entry for several weeks after generic hit the shelf. Channel inventory data provides the early warning.
3. Promotional impact: When evaluating whether a sales campaign drove prescription growth, always account for the lag. The effect of a campaign in week1 may not be measurable until week 3–5 of data.
4. Year-end data: December data is often restated in January as late-reporting pharmacies transmit their year-end records. Do not finalise year-end metrics until data has fully settled, typically 4–6 weeks into the new year.

The restatement problem — data you trusted last week may have changed:

Prescription data is not final when first delivered. As late-transmitting pharmacies submit their records, prior week’s data is restated, sometime 1%–3%, occasionally more. This means a weekly metric you reported 2 weeks ago may look different today as the data has settled. Best practice: add a data maturity indicator to dashboards, flag data from the most recent 2–3 weeks as preliminary, and avoid over-indexing on very recent data points for strategic decisions.

What Prescription Data Does Not Capture

Understanding the limitations of any data source is as important as understanding its contents. Here is what Rx data misses:

1. Unfilled prescriptions: If a physician write a prescription that the patient never fills, due to cost, access barriers, or patient preference, it never appears in Rx data. For speciality drugs with prior authorization requirements, the gap between prescription written and prescritpions filled can be substantial.
2. Hospital & Clinical dispensing: Drugs dispensed directly in hospitals or clinics, particularly IV infusions and injectables often do not flow through retail pharmacies and are not captured in standard NPA data. Medical claims data or hospital specific data sources are needed for these channels.
3. Samples: Drugs samples provided directly to physicians for patient use do not flow through pharmacies and are not counted in Rx data. Sample volume can be significant during early launch phases.
4. 340B dispensing: Drugs dispensed under the 340B federal drug discount programme have historically been under reported or inconsistently captured in commercial Rx data. This is an active area of data quality concern.
5. True patient identity: Because patient information is de-identified, standard Rx data cannot follow an individual patient longitudinally across refills or across drugs. Longitudinal patient level analytics requires a separate data product(will cover later).

Key Takeaways for Data Professionals

1. Prescription data captures dispensed prescriptions (drugs actually given to patients) not prescription written. The gap between written & filled is analytically significant for speciality drugs.
2. TRx, NRx & NBRx measure different things. TRx is total volume. NRx is new patient acquisition. NRRx is brand switching capture. Using the wrong metric to answer a business question leads to wrong conclusions.
3. All national Rx data involves statistical projection. No aggregator captures 100% of pharmacies. IQVIA’s projection is generally reliable at national & regional level but less so at small geographies or in speciality channels.
4. The 2–4 week data lag is structural & unavoidable. Use 852 wholesaler data as an early proxy for launch & LOE tracking. Account for the lag in all promotional impact analysis.
5. Rx data is restated as late transmitting pharmacies report. Flag recent data as preliminary & avoid over-indexing on the most recent week or two for strategic decisions.