Prescribing reasons, adherence barriers, perception of a launch: the AI proposes the codebook, codes every response, and your team reviews the doubtful ones before delivery.
Physician answers are short and technical; patient answers are long and emotional. Both have to end up in a clear table.
A physician writes "safety profile, once-daily dosing, covered by formulary" in one line: three different reasons in one answer.
Patients mix cost, side effects they feel and forgetting in one sentence about why they stopped treatment.
A comment may mention a possible adverse event, and the team needs to find it for its own pharmacovigilance process.
Multi-country studies arrive in several languages and the client expects one codebook to compare markets.
The AI reads all the responses and proposes a codebook, or applies the one the study already has. Every response gets all the codes that apply, and doubtful ones go to human review.
Nets that separate efficacy, safety, dosing, access and cost in prescribing reasons.
You can create a code that flags mentions of possible adverse events and filter them so your team reviews them under its own procedure.
Responses from several countries and languages coded with the same codebook.
Codes as columns per respondent in the .sav or Excel file, and tabulation with specialty or country in the banner.
Applications for pharmaceutical research
Mapping the patient journey from diagnosis to treatment.
Analysis of physician opinions on treatments and products.
Coding open comments from participants in observational and patient experience studies.
New medication and treatment acceptance analysis.
Treatment adherence barriers and drivers.
Qualitative data coding from observational studies.
Survey of patients on a chronic treatment who report having stopped or skipped doses. Here is part of the dataset after coding.
Open-ended question
Why did you stop taking the medication, or why do you sometimes skip doses?
| Response | Codes |
|---|---|
| “I forget the evening one, especially on weekends.” | Forgetting |
| “My pharmacy is often out of stock and paying out of pocket elsewhere is really expensive.” | Out of stockMedication cost |
| “It made me dizzy in the mornings so I decided to stop for a while.” | Perceived discomfortPossible adverse event (review) |
| “I feel fine now so I figured I didn't need it anymore.” | Feels well / thinks it's no longer needed |
| “Too many pills a day, between this one and my blood pressure meds I get mixed up.” | Too many doses a day |
| “My doctor switched me to this and didn't really explain how to take it.” | Not enough information from doctor |
| “My follow-up appointment was three months away and my prescription ran out.” | Appointment access / prescription expired |
| Net / code | % of respondents who answered |
|---|---|
| NET Access and cost | 36% |
| Medication cost | 17% |
| Out of stock | 14% |
| Appointment access / prescription expired | 11% |
| NET Routine | 34% |
| Forgetting | 26% |
| Too many doses a day | 12% |
| NET Treatment perception | 30% |
| Feels well / thinks it's no longer needed | 16% |
| Perceived discomfort | 15% |
| Not enough information from doctor | 6% |
| NET For team review | 9% |
| Possible adverse event (review) | 9% |
Multi-code question: one response can carry several codes, so percentages add up to more than 100%. A net counts each respondent once, even if they have several codes from that group.
Example built for this page: this is not client data and the percentages are illustrative.
No. It does not file reports or replace the pharmacovigilance process. What you can do is keep a code in the codebook to flag comments that mention a possible adverse event, filter them and hand them to your team, which applies its own procedure.
It is not a specialized medical model. It codes from what the responses say and from the codebook you generate or load. That's why doubtful answers go to human review, and your team can adjust codes to the terminology the study uses.
Yes. The AI codes them against the codebook, and the ones that aren't clear are flagged as doubtful. It helps for the codebook to have explicit code definitions (for example, separating "dosing" from "route of administration").
Yes. Responses in different languages are coded with the same codebook, so the table by country uses the same categories.
Upload a dataset with an open-ended question and see the codebook the AI proposes. The free trial includes 250 responses.