UCLA Health Patient Health Maintenance Text Reminder

This is a prospective randomized clinical trial evaluating how behaviorally-informed outreach text message reminders impact patient engagement with primary care. This prospective randomized control trial is being implemented in conjunction with UCLA Health's larger quality improvement initiative (the My Action Plan Quality Improvement Initiative) in order to improve primary care preventive measure completion rates.

UCLA Health will launch an initiative called My Action Plan which is an outreach effort targeting the primary care patients at UCLA who have an outstanding, clinically indicated primary care preventative care gap (e.g., overdue colorectal cancer screening, outstanding diabetes tests) and aims at encouraging patients to complete their overdue health maintenance measures.

At the beginning of each month, identified UCLA Health primary care patients will be enrolled in a trial the investigators have pre-registered under the title "UCLA Health Patient Health Maintenance Outreach Text Message" and randomized to receive one of four text messages informing them that they have overdue health maintenance measures and urging them to take action to complete these measures. These text messages will contain a link to their MyChart patient portal account, specifically linking them to a My Action Plan electronic letter that will contain a personalized list of outstanding preventive care items and actionable steps to complete the items.

In March and April 2023, if participants do not open their My Action Plan electronic letter within 7 days after the initial outreach, they will be enrolled in the trial described in the current pre-registration. This trial will evaluate the effect of sending a text reminder about health gaps as well as compare the effectiveness of different outreach text reminders. Patients will be randomly assigned in a 1:1:1:1 ratio to either the holdout control condition (without a reminder) or one of the three reminder arms (receiving one of three text reminders 2 weeks after the initial outreach message). Additionally, within each text-reminder arm, patients will be randomized to one of three time slots for getting the reminder. The investigators will examine patients' interaction with these reminders and how that ultimately translates to them completing their indicated health maintenance measures.

The UCLA Department of Medicine (DOM) intends to roll out this QI initiative at the beginning of the 2023 calendar year and continue it for a span of at least 12 months. Each month, the initiative targets patients whose birthday falls into that month. This proposed trial will run for two months (March and April 2023).

Analysis Plan:

The investigators will utilize patient-level ordinary least squares (OLS) estimation, with statistical inferences based on model-robust standard errors. The primary model term will be indicator variables for arm assignment.

The analysis will adjust for sex, age, race/ethnicity, indicators for screenings/tests that patients are due for, and whether patients have upcoming primary care appointments. Missing covariate values will be handled by including 'unknown' indicators, along with mean imputation for quantitative covariates.

The investigators will use this dataset to address two separate research questions.

Research question 1: The investigators will compare the average of Arms 2-4 with Arm 1 to test the average effect of sending a text reminder. If the average effect is significant, the investigators will explore if Arms 2, 3, and 4 separately differ from Arm 1.

Research question 2: The investigators will compare Arm 3, Arm 4, with Arm 2 to test whether, compared to a simple (control) reminder, it is more effective to boost patients' intentions to close the health gap via an anecdote or to prompt patients to take actions quickly.

Exploratory analyses will investigate heterogeneous treatment effects by patients' baseline motivation (i.e., the extent to which patients seem to have some intentions to get the screenings/tests) in two ways. First, the investigators will use each patient's history of screenings/tests to construct a proxy for their baseline motivation to obtain the due screenings/tests. Specifically, the investigator will calculate, among all the screenings/tests that are included in the My Action Plan Initiative and that a patient was due for in the past few years (exact time window TBD based on data availability), what percentage was completed by the patient (as far as UCLA Health could tell). A higher percentage indicates a higher baseline motivation to get the screenings/tests patients are due for during the study period. Second, the investigators will obtain demographics and medical information that UCLA Health is willing to provide (e.g., gender, age, race/ethnicity, history of cancer and other pre-existing conditions, family cancer history, past adherence to screenings/tests that are included in the My Action Plan Initiative, past receipt of influenza vaccination, frequency of doctor visits; time window of these variables is TBD depending on data availability). With such information, the investigators will train an algorithm to predict patients' baseline motivation level using patients in the holdout condition (Arm 1) as the ground truth. Specifically, the investigators will take the aforementioned information about patients in Arm 1 as input, and use whether they complete any of the screenings/visits recommended in the My Action Plan letter within 6 months as the outcome measure. Then the investigators will apply the algorithm to all patients to predict their baseline motivation level. For both approaches, the proxy for baseline motivation (Approach 1) and the predicted baseline motivation level (Approach 2) will be used to analyze the heterogeneous treatment effects of (1) Arms 3 and 4 (vs. Arm 2) and (2) the combination of Arms 2 and 4 vs. Arm 1.

Additionally the investigators will investigate proxies for whether patients face structural barriers to get screenings/tests as moderators, including socioeconomic factors at the zipcode level, insurance type, proxies for accessibility to healthcare resources such as distance from UCLA Health clinics.

The investigators will also explore how the effect of text reminders varies across the three times of the day when the reminders will be sent.