Digital Gradebook Engagement Scores May Flag Adolescents at Risk for Substance Use

Engagement scores that teachers enter into digital gradebooks, covering classroom behavior and participation, distinguished middle school students at high and low risk of substance use better than their grade point average (GPA), according to a secondary analysis of a randomized controlled trial in Los Angeles, California.¹

The investigators sought to find a passive, low-cost way to spot adolescents at risk of substance use. Many adolescents may be reluctant to tell adults about risky behaviors and only about one third have an annual well-child visit. The investigators believe academic data could be a possible behavioral health “vital sign.”

Study Design

The analysis drew on data from the Linking Information and Families Together (LIFT) trial (NCT02129153).^1,2 In that trial, parents of seventh graders at 4 public middle schools received weekly texts, calls, or emails about missed assignments, grades, and behavior. Participants were recruited in sixth grade and followed through eighth grade. They came mostly from low-income households.

Students completed computer-assisted surveys based on Youth Risk Behavior Surveillance Survey measures. The composite outcome was any self-reported lifetime use of alcohol or cannabis by eighth grade. Other illicit drug use was close to zero and was left out of the outcome.

The predictors fell into 2 groups. The first was GPA, both transcript-recorded and parent-reported. The second was digital gradebook data in 3 domains: summative (quizzes and tests), formative (homework), and engagement (classroom behavior and participation). The investigators also built a trajectory variable for change in gradebook scores from seventh to eighth grade. They checked discriminative performance using k-fold cross-validation.

The complete-case sample had 234 students. Of these, 51.7% were female and 82.1% were Hispanic/Latino, and 47 reported the composite outcome. One school did not record engagement data, so engagement models used a subgroup of 170 students.¹

Gradebook vs GPA

In unadjusted analyses, substance use was significantly associated with formative scores (OR=0.95; P=.025), engagement scores (OR=0.86; P<0.001), and a low-stable gradebook trajectory (OR=1.95; P=.044). A high-stable trajectory was protective (OR=0.46; P=.038). Parent-reported GPA in seventh grade was also associated with the outcome (OR=0.65; P=.026); Transcript-recorded GPA in sixth or seventh grade was not. After adjustment, engagement score was the strongest and most consistent predictor (OR=0.84; P<0.001).¹

The engagement model had a c-statistic of 0.80 and a cross-validated AUC of 0.79 (95% CI, 0.683-0.871). Adding parent-reported GPA did not improve it: the cross-validated AUC fell to 0.74 (95% CI, 0.614-0.856). GPA-only models did worse. The model using both GPA measures had a cross-validated AUC of 0.63, and the model using parent-reported GPA alone had 0.59. The formative model reached 0.60. A sensitivity analysis using multiple imputation across 40 datasets gave very similar estimates.

The authors suggest that formative and engagement measures may pick up behavioral, relational, and environmental influences, such as student-teacher conflict or a home environment that makes studying hard. These influences may be more closely tied to substance use risk than academic mastery is. The investigators did not test these mechanisms directly.

Implementation and Equity

The authors frame gradebook data as a passive monitoring system. It requires no work beyond teachers’ routine reporting and does not depend on student self-report. That sets it apart from clinic-based tools such as the Pediatric Symptom Checklist, which miss youth with inconsistent access to health care.

They also warn that gradebook scores can reflect classroom management, grading culture, and implicit bias. Prediction models built on these data could amplify inequities for marginalized groups. They call for transparent models, monitoring of performance across populations, and compliance with the Family Educational Rights and Privacy Act. They add that any flag should lead to supportive conversations and connection to resources, not punitive action. In their view, such a system would serve as one input for multidisciplinary school support teams, not as a stand-alone screening tool.

The study had several limitations. The sample was small and came from 1 metropolitan area, and the analysis was complete-case. Because 1 school recorded no engagement data, the authors note that local validation and calibration will likely be necessary wherever the approach is used. Future research should test whether gradebook data predict other outcomes, including substance use disorders, mental health disorders, and violence. It should also test the approach in rural, suburban, and socioeconomically diverse populations.

References

1. DiPentino R, Meza BPL, Jackson N, et al. Could school function be an adolescent health vital sign? Associations between gradebook data and substance use among middle school students. J Sch Health. 2026;96(10):e70261.

2. Linking Families Together Study- A Randomized Trial to Raise Parental Monitoring (LIFT). Clinicaltrials.gov. Accessed November 12, 2026. https://clinicaltrials.gov/study/NCT02129153