AI adoption in primary care

Last registered on October 31, 2025

Pre-Trial

Trial Information

General Information

Title
AI adoption in primary care
RCT ID
AEARCTR-0017116
Initial registration date
October 27, 2025

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
October 31, 2025, 8:25 AM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Region

Primary Investigator

Affiliation
WZB & DIW Berlin

Other Primary Investigator(s)

PI Affiliation
University of Copenhagen & DIW Berlin
PI Affiliation
Stockholm School of Economics & University of Copenhagen

Additional Trial Information

Status
In development
Start date
2025-10-29
End date
2025-12-31
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
This study examines how physicians make antibiotic prescribing decisions and their use of diagnostic information. Using a vignette experiment with hypothetical patient cases, we aim to understand the antibiotic prescribing decisions of active primary care physicians. In between-subject treatments, we assess the impact of hypothetical AI support tools compared to traditional dipstick tests on physicians' diagnoses and treatment decisions. By matching the survey data with administrative data, we explore heterogeneity in treatment effects by real-life prescription behavior. Moreover, we study in how far physicians take the correlation between different diagnostic signals into account.
External Link(s)

Registration Citation

Citation
Huang, Shan, Renke Schmacker and Hannes Ullrich. 2025. "AI adoption in primary care." AEA RCT Registry. October 31. https://doi.org/10.1257/rct.17116-1.0
Experimental Details

Interventions

Intervention(s)
Vignettes 1-6:
AI tool vs dipstick: In the first treatment arm, we vary whether the diagnostic information is based on an AI support tool or a dipstick test, keeping the informativeness of the test fixed.

Vignettes 7-9:
Correlated vs. uncorrelated signals: In the second treatment arm, we vary whether the AI support tool includes the dipstick test or not.
Intervention (Hidden)
Intervention Start Date
2025-10-29
Intervention End Date
2025-12-31

Primary Outcomes

Primary Outcomes (end points)
- Belief updating regarding UTI probability
- Antibiotic prescriptions
- Prescription threshold
- Laboratory testing of urine sample
Primary Outcomes (explanation)
Belief updating: We elicit respondents’ UTI beliefs before receiving a signal (AI tool or dipstick) and after receiving the signal. We will use the average absolute difference between prior and posterior as a primary outcome. Additionally, we will use regressions to determine the movement in beliefs.

Antibiotic prescriptions: In the vignettes, we ask respondents whether they would prescribe antibiotics immediately, make a wait-and-see prescription, or not prescribe antibiotics.

Prescription threshold: For each respondent, we use the within-respondent variation in beliefs and prescription decisions to determine the lowest probability (threshold) probability at which they prescribe antibiotics.

Laboratory testing: In the vignettes, we ask whether respondents would seek further information by sending a sample to the lab or making a culture.

Secondary Outcomes

Secondary Outcomes (end points)
Guideline appropriateness
Secondary Outcomes (explanation)
Guideline appropriateness: We check whether prescriptions are consistent with the medical guidelines of DSAM (prescription and days of antibiotic intake).

Experimental Design

Experimental Design
See below
Experimental Design Details
We will employ a vignette survey experiment to assess treatment decisions by presenting physicians with randomized patient scenarios. Each vignette represents a hypothetical patient case, including background information and symptoms. First, we elicit the prior disease probability based on the patient characteristics and the prescription decision. On the next page, physicians observe a randomly drawn signal from either an AI support tool (AI tool treatment) or from a dipstick test (Dipstick treatment). After seeing the signals, we elicit their posterior disease probability again, as well as their final treatment decision. This choice situation is repeated for six vignettes. Afterwards, participants see another three vignettes, but this time they receive both the AI support tool and the dipstick signal. We vary experimentally whether the dipstick signal is included in the AI signal (Correlated treatment) or whether it is not included (Uncorrelated treatment). Again, we ask for their diagnostic beliefs and treatment decisions.

We test the following experimental hypotheses:

H1: Physicians update differently to the AI signal compared to the dipstick signal.
H2: Physicians update differently in the Correlated treatment compared to the Uncorrelated treatment.

We will study heterogeneous treatment effects for hypotheses H1 and H2 by age, gender, education (university attended), prescription intensity (based on admin data; individual and clinic-level prescribing), clinic characteristics (such as clinic size, number of physicians working at the clinic, consultations per physician, rural/urban, based on admin data), diagnostic intensity (based on admin data; individual and clinic-level claims), and time preferences (based on survey data).

To test the hypotheses, we will compare the average absolute updating using parametric and non-parametric tests. Moreover, we will estimate Bayesian updating regressions following Grether (1992, JEBO) and compare the updating coefficients. In further analyses, we will compare updating to positive vs. negative signals.

After the vignettes, we ask for physician attitudes towards the AI support tool and the dipstick test. We will analyze whether these attitudes predict average belief updating to the respective signal. Moreover, we study whether general attitudes toward AI predict belief updating to the AI signal.

We will exploratively investigate whether differences in prescription and diagnostic intensity (in the survey and the admin data) are associated with differences in prior beliefs, reaction to signals, and/or different prescription thresholds (as measured in the experiment). Similarly, we will investigate whether individuals with different concern about AMR externality differ with respect to these factors.

Moreover, we elicit a range of physician characteristics (risk, time, and social preferences, concern about antimicrobial resistance, belief about own prescription intensity) and will exploratively analyze whether these characteristics predict antibiotic prescription decisions in the survey as well as prescribing and diagnostic intensities in the administrative data. Furthermore, we will exploratively analyze the determinants of technological and AI adoption.

For robustness, we will exclude participants who repeatedly updated in the wrong direction (i.e., who updated positively to a negative signal, or negatively to a positive signal).
Randomization Method
Randomization by a computer
Randomization Unit
Individual
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
Target sample: 500 subjects in total
We aim for a target sample of 500 subjects, but the final sample size will depend on the response rate to our invitations.
Sample size: planned number of observations
Target sample: 500 subjects in total We aim for a target sample of 500 subjects, but the final sample size will depend on the response rate to our invitations.
Sample size (or number of clusters) by treatment arms
AI tool treatment: 1/2 of sample
Dipstick treatment: 1/2 of sample
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
H1: Assuming that subjects update on average 17 ppts (sd=10) from prior to posterior after seeing a dipstick signal, with 250 subjects per treatment arm, we can detect a change of 2.5 ppts with 80% power at the p=0.05 level. H2: Assuming that subjects update on average 21 ppts (sd=14) from prior to posterior after seeing a correlated signal, with 250 subjects per treatment arm, we can detect a change of 3.5 ppts with 80% power at the p=0.05 level.
IRB

Institutional Review Boards (IRBs)

IRB Name
Research Ethics Committee, Department of Economics, University of Copenhagen
IRB Approval Date
2024-06-06
IRB Approval Number
N/A

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials