Timely patient arrival is essential for maintaining efficient clinic operations and ensuring patients receive the full duration of their scheduled care. Late arrivals disrupt clinic flow, reduce time with providers, and can delay subsequent appointments, creating a ripple effect across the day. This project aims to test a low-cost behavioral intervention that clarifies arrival expectations and promotes on-time arrival for scheduled appointments.
This study evaluates the effectiveness of an intervention designed to make the expected arrival time more salient to patients. The study will be implemented across outpatient departments that deliver primary care as part of UCLA Health's operational improvement initiative to improve clinic flow and patient experience.
Specifically, the intervention modifies how appointment information is presented to patients (or proxies in the case of patients \< 18 years old) through two communication channels used by UCLA Health: (1) the patient portal (called MyChart at UCLA Health) and (2) the text messaging platform (called Hello World at UCLA Health), which sends patients automated text notifications prior to their appointment.
Control Condition On the MyChart home page, patients see only the scheduled appointment time (e.g., "Starts at 3:00 PM") for their upcoming appointments. Within the appointment details page, patients see their appointment time and an additional instruction to "Arrive 15 minutes before your appointment." Text notifications state the appointment time and include a prompt to arrive 15 minutes early (e.g., "\[Patient name\], you have an upcoming visit on 12 Jan 2026 at 3:00pm. Arrive 15 minutes before your appointment time").
Treatment Condition On the MyChart home page, patients see an explicit arrival time for their upcoming appointments, set to 15 minutes before the scheduled appointment time (e.g., "Arrive by 2:45 PM" for an appointment scheduled at 3PM), instead of the appointment time itself. Within the appointment details page, patients see their appointment time as well as the arrival time.
Text notifications explicitly tell patients their expected arrival time, instead of their appointment time (e.g., "\[Patient name\], you have an upcoming visit on 12 Jan 2026. Arrive by 2:45pm.")
The investigators will implement a cluster randomization, assigning all in-person physician or advanced practice provider (APP) appointments within the same clinic building to the same experimental condition. A total of 37 clinic buildings will be included. The investigators plan to run this clinical trial for 6 months. The arrival time functionality will be implemented 1 month before the start of the trial period to introduce a washout period.
Key Research Question:
The investigators ask the following research question: Does displaying an explicit arrival time change how early patients arrive for their appointments? The investigators hypothesize that the treatment will lead people to arrive earlier for their appointments, relative to the control condition.
In addition to examining the continuous measure of how far ahead a patient checks in for their appointments (the primary outcome measures), the investigators will also examine two binary metrics:
A. On-time arrival: Does the treatment lead patients to be more likely to arrive on time (i.e., check in before or at the scheduled appointment time)?.
B. Early arrival: Does the treatment condition increase patients' likelihood of arriving at least 15 minutes before their scheduled appointment time?
Analysis Plan:
The investigators will use ordinary least squares (OLS) regressions with robust standard errors to predict outcome variables. Statistical inferences will be based on model-robust standard errors clustered at the clinic building level, which is the unit of randomization. All analyses will be conducted at the appointment level. The primary predictor will be an indicator for whether the patient's clinic was assigned to the arrival time treatment condition (vs. standard appointment time control).
These regressions will be run with and without the following control variables:
* Patient age (If there will be missing values, the investigators will replace with the mean and add a dummy variable to indicate patients with missing age)
* Patient gender
* Patient race/ethnicity
* Patient punctuality: The proportion of the patient's in-person physician appointments at the same clinic in the past 36 months prior to the start of the trial for which they arrived on time (Patients with no completed appointments during this period will have missing data. For these cases, the investigators will impute the sample mean and add a dummy variable to indicate these patients). If we cannot have patient visit history for 36 months prior to the start of the trial, we will instead control for the proportion of the patient's in-person physician appointment at UCLA Health (across all clinics) for which they arrived on time (excluding walk-in appointment)
* Visit type (adult vs child; new vs. return)
* An indicator for whether the clinic operated using appointment schedules in 15/30 minute slots or 20/40 minute slots
* Insurance type (Traditional Medicare, Medicare Advantage, Medicaid, Commercial, other/unknown)
* Multimorbidity score (If there are missing values , the investigators will replace with the mean and add a dummy variable to indicate patients with a missing multimorbidity score)
* The following mutually exclusive binary indicators to account for differences in (1) whether an appointment has been scheduled before the arrival-time functionality was implemented and (2) how far in advance an appointment was scheduled :
* Scheduled before functionality implementation: Indicator for appointments that were scheduled before the arrival-time functionality was implemented
* Scheduled \>7 days in advance: Indicator for appointments scheduled on or after the implementation date with more than 7 days between scheduling and appointment date.
* Scheduled 1-7 days in advance: Indicator for appointments scheduled on or after the implementation date with 1 to 7 days between scheduling and appointment date.
* Scheduled \<1 day in advance: Indicator for appointments scheduled on or after the implementation date with less than 1 day between scheduling and appointment date.
As an exploratory analysis, the investigators will examine treatment effects only among appointments scheduled on or after the functionality implementation date, since the appointment confirmation text notifications received by those patients will reflect the experimental design.
For robustness checks, the investigators will conduct logit models for binary outcomes and re-estimate models for the continuous arrival-time outcome at the 1st and 99th percentiles (which is a wider winsorization range than the range specified in the primary outcome section later).
Note: Some patients may have multiple eligible encounters during the study window. The main analyses will include only the first encounter per patient. Exploratory analyses will assess whether including subsequent encounters changes the results.
To test for heterogeneous treatment effects, the investigators will estimate interaction models in which the treatment indicator is interacted with patient, visit, or clinic characteristics. The investigators will analyze the following moderators:
* Patient-level punctuality as defined above (To account for missing values, the regression models will interact the treatment indicator with both the imputed punctuality rate and the missingness indicator)
* How many times patients have attended a physician appointment at that same clinic in the 36 months prior to the start of the trial
* Whether the clinic department schedules appointments in 15/30 minute slots or 20/40 minute slots
* Clinic's baseline punctuality rate: percentage of appointments that were on time (checked in ≤ 0 minutes late) in the six months prior to the trial
* Visit type (adult vs. child)
* Appointment scheduling time (four mutually exclusive subgroups): Whether the appointment was scheduled before the arrival-time functionality was implemented vs. it was scheduled after the functionality was implemented but more than 7 days before the scheduled visit date vs. 1-7 days, vs. \<1 day before the scheduled visit date.
* Multimorbidity score (To account for missing values, the regression models will interact the treatment indicator with both the imputed multimorbidity score and the missingness indicator)
The investigators will conduct additional exploratory analyses to examine potential downstream consequences of the intervention:
1. Appointment attendance The investigators will explore whether the treatment condition affects no-show rates relative to the control condition. Explicitly displaying an earlier requested arrival time could potentially increase no-shows if some patients believe they have already missed their appointment window or feel discouraged from coming when they realize they cannot arrive by the requested time.
2. Patient satisfaction:
The investigators will explore whether assignment to the treatment condition affects patient satisfaction ratings in post-visit surveys, by focusing on two binary outcomes that allow for an intent-to-treat analysis:
A. Whether the patient leaves a bad review (this variable equals 1 if the patient gives 1 or 2 stars out of 5 stars and 0 if the patient gives a higher star rating or does not provide a star rating) B. Whether the patient leaves a good review (this variable equals 1 if the patient gives 5 stars and 0 if the patient gives a lower rating or does not provide a star rating).
The investigators will conduct text analyses of patients' qualitative comments in the surveys to assess sentiment and sources of concerns or compliments.
3. Appointment cancellations The investigators will test whether the treatment condition affects cancellation rates relative to the control condition. This analysis will help determine whether the "Arrive By" framing has any unintended effect on patients' likelihood of cancelling rather than attending their appointment.
4. Clinic workflow metrics:
The investigators will examine whether the treatment influences the following two operational metrics within the clinic:
A. Patient room time: the number of minutes between the patient's check-in and the time they are placed in the exam room.
B. Time when physician enters the exam room: the number of minutes between the scheduled appointment time and when the provider enters the room.
C. Time between check-in and provider entry: Number of minutes between the patient's check-in and the time the provider enters the exam room.
The investigators will use quantile regressions for this set of analyses. These exploratory measures provide insight into whether improved punctuality translates into smoother clinic flow or changes in patient wait time. These measures are likely noisy as the time when patients are placed in the exam room or are seen by the provider are not always recorded in real time.
