Trial Outcomes & Findings for A Trial of AI-Powered Text Message Outreach on Well-Child Visit Completion (NCT NCT06698640)

NCT ID: NCT06698640

Last Updated: 2026-06-22

Results Overview

Binary indicator of whether each participant completed at least one well-child visit by December 31, 2025, ascertained through administrative claims using HEDIS technical specifications

Recruitment status

COMPLETED

Study phase

NA

Target enrollment

2821 participants

Primary outcome timeframe

Up to 7 months post-randomization (June 1 - December 31, 2025)

Results posted on

2026-06-22

Participant Flow

Of 3,908 Medicaid beneficiaries meeting eligibility, 3,071 were overdue for a well-child visit per HEDIS. Of these, 2,847 were attributed to the trial's Virginia health system (224 out-of-network not randomized). After excluding 26 (do-not-contact n=19; inactive phone n=7), 2,821 participants in 2,039 households were randomized 1:1:1 on June 1, 2025 (Arm 1: 618 households/927 children; Arm 2: 633/949; Arm 3: 788/945).

Only Medicaid-enrolled children aged 0-21 years were enrolled. Parents/caregivers who received texts or interacted with the AI scheduler on behalf of children were not enrolled; no parent/caregiver data were collected and they are not represented in Baseline Characteristics, Outcome Measures, or Adverse Events. Randomization was conducted at the household level on June 1, 2025, in a 1:1:1 ratio.

Unit of analysis: Household

Participant milestones

Participant milestones
Measure
Control
Participants received standard health plan outreach consisting of periodic mailed reminders. Families retained access to all standard appointment scheduling methods including telephone calls to provider offices.
Automated SMS
Participants received standardized automated text message reminders at predetermined intervals (initial contact, 2-week follow-up, 4-week follow-up). Messages provided general information about visit importance and instructions to call providers to schedule. Messages did not include interactive scheduling or conversational capabilities.
Automated SMS + Scheduling Assistance
Participants received automated text message reminders with response options. When families expressed interest, structured follow-up texts collected appointment preferences. An AI-powered telephone scheduling system (GPT-4o, OpenAI) contacted the child's primary care provider directly to book appointments on behalf of families. The system disclosed its automated nature at call initiation and escalated to human staff when clinic policies were incompatible with automated scheduling or technical failures occurred.
Overall Study
STARTED
927 618
949 633
945 788
Overall Study
COMPLETED
927 618
949 633
945 788
Overall Study
NOT COMPLETED
0 0
0 0
0 0

Reasons for withdrawal

Withdrawal data not reported

Baseline Characteristics

A Trial of AI-Powered Text Message Outreach on Well-Child Visit Completion

Baseline characteristics by cohort

Baseline characteristics by cohort
Measure
Automated SMS
n=949 Participants
Participants received standardized automated text message reminders at predetermined intervals (initial contact, 2-week follow-up, 4-week follow-up). Messages provided general information about visit importance and instructions to call providers to schedule. Messages did not include interactive scheduling or conversational capabilities.
Control
n=927 Participants
Participants received standard health plan outreach consisting of periodic mailed reminders. Families retained access to all standard appointment scheduling methods including telephone calls to provider offices.
Automated SMS + Scheduling Assistance
n=945 Participants
Participants received automated text message reminders with response options. When families expressed interest, structured follow-up texts collected appointment preferences. An AI-powered telephone scheduling system (GPT-4o, OpenAI) contacted the child's primary care provider directly to book appointments on behalf of families. The system disclosed its automated nature at call initiation and escalated to human staff when clinic policies were incompatible with automated scheduling or technical failures occurred.
Total
n=2821 Participants
Total of all reporting groups
Age, Continuous
13.0 Years
STANDARD_DEVIATION 5.4 • n=20 Participants
12.8 Years
STANDARD_DEVIATION 5.4 • n=20 Participants
12.7 Years
STANDARD_DEVIATION 5.3 • n=40 Participants
12.8 Years
STANDARD_DEVIATION 5.4 • n=6 Participants
Sex: Female, Male
Female
474 Participants
n=20 Participants
449 Participants
n=20 Participants
455 Participants
n=40 Participants
1378 Participants
n=6 Participants
Sex: Female, Male
Male
475 Participants
n=20 Participants
478 Participants
n=20 Participants
490 Participants
n=40 Participants
1443 Participants
n=6 Participants
Race (NIH/OMB)
American Indian or Alaska Native
0 Participants
n=20 Participants
0 Participants
n=20 Participants
0 Participants
n=40 Participants
0 Participants
n=6 Participants
Race (NIH/OMB)
Asian
41 Participants
n=20 Participants
70 Participants
n=20 Participants
56 Participants
n=40 Participants
167 Participants
n=6 Participants
Race (NIH/OMB)
Native Hawaiian or Other Pacific Islander
0 Participants
n=20 Participants
0 Participants
n=20 Participants
0 Participants
n=40 Participants
0 Participants
n=6 Participants
Race (NIH/OMB)
Black or African American
379 Participants
n=20 Participants
386 Participants
n=20 Participants
378 Participants
n=40 Participants
1143 Participants
n=6 Participants
Race (NIH/OMB)
White
331 Participants
n=20 Participants
303 Participants
n=20 Participants
330 Participants
n=40 Participants
964 Participants
n=6 Participants
Race (NIH/OMB)
More than one race
11 Participants
n=20 Participants
2 Participants
n=20 Participants
6 Participants
n=40 Participants
19 Participants
n=6 Participants
Race (NIH/OMB)
Unknown or Not Reported
187 Participants
n=20 Participants
166 Participants
n=20 Participants
175 Participants
n=40 Participants
528 Participants
n=6 Participants
Ethnicity (NIH/OMB)
Hispanic or Latino
1 Participants
n=20 Participants
3 Participants
n=20 Participants
6 Participants
n=40 Participants
10 Participants
n=6 Participants
Ethnicity (NIH/OMB)
Not Hispanic or Latino
762 Participants
n=20 Participants
761 Participants
n=20 Participants
770 Participants
n=40 Participants
2293 Participants
n=6 Participants
Ethnicity (NIH/OMB)
Unknown or Not Reported
186 Participants
n=20 Participants
163 Participants
n=20 Participants
169 Participants
n=40 Participants
518 Participants
n=6 Participants
Number of Participants with Open Well-Child Visit Gap at Baseline
949 Participants
n=20 Participants
927 Participants
n=20 Participants
945 Participants
n=40 Participants
2821 Participants
n=6 Participants

PRIMARY outcome

Timeframe: Up to 7 months post-randomization (June 1 - December 31, 2025)

Population: All randomized participants with available claims data through December 31, 2025 (intent-to-treat population). Participants with no claims data were assumed to have not completed a well-child visit.

Binary indicator of whether each participant completed at least one well-child visit by December 31, 2025, ascertained through administrative claims using HEDIS technical specifications

Outcome measures

Outcome measures
Measure
Control
n=927 Participants
Participants received standard health plan outreach consisting of periodic mailed reminders. Families retained access to all standard appointment scheduling methods including telephone calls to provider offices.
Automated SMS
n=949 Participants
Participants received standardized automated text message reminders at predetermined intervals (initial contact, 2-week follow-up, 4-week follow-up). Messages provided general information about visit importance and instructions to call providers to schedule. Messages did not include interactive scheduling or conversational capabilities.
Automated SMS + Scheduling Assistance
n=945 Participants
Participants received automated text message reminders with response options. When families expressed interest, structured follow-up texts collected appointment preferences. An AI-powered telephone scheduling system (GPT-4o, OpenAI) contacted the child's primary care provider directly to book appointments on behalf of families. The system disclosed its automated nature at call initiation and escalated to human staff when clinic policies were incompatible with automated scheduling or technical failures occurred.
Number of Participants With Well-Child Visit Completion
218 Participants
209 Participants
280 Participants

SECONDARY outcome

Timeframe: Up to 7 months post-randomization (June 1 - December 31, 2025)

Population: Time-motion analysis at the appointment level. For traditional scheduling (Arm 1), 100 randomly sampled appointments scheduled manually by care team staff were analyzed using EHR timestamps, Twilio call logs, and activity logs. For AI-facilitated scheduling (Arm 3), all 63 successfully booked appointments were analyzed. Arm 2 contributed no appointments as that arm received reminders only without scheduling assistance. In both analyzed arms, each appointment corresponded to a unique participant.

Mean staff time in minutes per successfully scheduled appointment, assessed via time-motion analysis. For traditional care team scheduling, staff time was measured from initial scheduling activity to appointment confirmation using EHR timestamps, Twilio call logs, and activity logs. For AI-facilitated scheduling, staff time was measured from appointment request to completion of quality assurance review.

Outcome measures

Outcome measures
Measure
Control
n=100 Scheduled appointment
Participants received standard health plan outreach consisting of periodic mailed reminders. Families retained access to all standard appointment scheduling methods including telephone calls to provider offices.
Automated SMS
Participants received standardized automated text message reminders at predetermined intervals (initial contact, 2-week follow-up, 4-week follow-up). Messages provided general information about visit importance and instructions to call providers to schedule. Messages did not include interactive scheduling or conversational capabilities.
Automated SMS + Scheduling Assistance
n=63 Scheduled appointment
Participants received automated text message reminders with response options. When families expressed interest, structured follow-up texts collected appointment preferences. An AI-powered telephone scheduling system (GPT-4o, OpenAI) contacted the child's primary care provider directly to book appointments on behalf of families. The system disclosed its automated nature at call initiation and escalated to human staff when clinic policies were incompatible with automated scheduling or technical failures occurred.
Staff Time Per Successfully Scheduled Appointment
14.0 Minutes
Standard Deviation 4.1
2.0 Minutes
Standard Deviation 1.0

SECONDARY outcome

Timeframe: Up to 2 weeks from initiation of first automated scheduling attempt

Population: All Arm 3 participants for whom at least one automated scheduling attempt was initiated during the intervention period (June 9 - July 14, 2025).

Percentage of automated scheduling attempts in Arm 3 requiring human staff intervention, with reasons for escalation documented.

Outcome measures

Outcome measures
Measure
Control
n=75 Scheduling attempts
Participants received standard health plan outreach consisting of periodic mailed reminders. Families retained access to all standard appointment scheduling methods including telephone calls to provider offices.
Automated SMS
Participants received standardized automated text message reminders at predetermined intervals (initial contact, 2-week follow-up, 4-week follow-up). Messages provided general information about visit importance and instructions to call providers to schedule. Messages did not include interactive scheduling or conversational capabilities.
Automated SMS + Scheduling Assistance
Participants received automated text message reminders with response options. When families expressed interest, structured follow-up texts collected appointment preferences. An AI-powered telephone scheduling system (GPT-4o, OpenAI) contacted the child's primary care provider directly to book appointments on behalf of families. The system disclosed its automated nature at call initiation and escalated to human staff when clinic policies were incompatible with automated scheduling or technical failures occurred.
Human Escalation Rate for Automated Scheduling Attempts
16.0 Percentage of scheduling attempts

Adverse Events

Control

Serious events: 0 serious events
Other events: 0 other events
Deaths: 0 deaths

Automated SMS

Serious events: 0 serious events
Other events: 0 other events
Deaths: 0 deaths

Automated SMS + Scheduling Assistance

Serious events: 0 serious events
Other events: 0 other events
Deaths: 0 deaths

Serious adverse events

Adverse event data not reported

Other adverse events

Adverse event data not reported

Additional Information

Sanjay Basu, MD PhD

Waymark

Phone: 415-577-5796

Results disclosure agreements

  • Principal investigator is a sponsor employee
  • Publication restrictions are in place