JMIR Publications

ABSTRACT

Background:

Remote electronic followup supports healthbehaviorchange interventions; nevertheless, suboptimal patient engagement and followup adherence remain key barriers for dizzinessheadache specialty care. Intelligentagentdriven followup represents a promising mHealth approach, yet realworld evidence for embedding these tools within clinical workflows remains limited.

Objective:

This study developed a fullcourse intelligent followup system for dizziness and headache care containing 5 functional modules: diseaserelated knowledge guidance, specialty diary, dailylife trigger management, medical consultation guidance, and data analysis. We aimed to evaluate the effects of agentassisted followup on followup adherence, platformmeasured patient engagement, and symptomrelated outcomes among patients experiencing recurrent dizziness or headache.

Methods:

An observational nonequivalent controlgroup, nonsynchronized prepost study design was adopted. Two hundred fiftytwo specialtyclinic patients (n=126 per group) were enrolled. The intervention group (AprilJuly 2026) received conventional electronic followup augmented with intelligentagent assistance. The historical control group (November 2025March 2026) received conventional electronic followup only. Primary outcomes included specialtydiary completion rate, monthly change in headache/dizziness days relative to baseline, symptomseverity metrics, and acutephase medication documentation. Secondary outcomes were platformderived patientengagement indicators. Frequencies and proportions summarized count data; the Wilcoxon ranksum (MannWhitney U) test was applied for nonnormallydistributed continuous variables. Bonferroni correction was implemented for multiple subgroup comparisons. All participants provided written informed consent; institutional ethics approval was obtained.

Results:

Specialtydiary completion rate was 72.2% (91/126) in the intervention group versus 52.4% (66/126) in controls (χ²=10.56, df=1, P=.001). In the migraine subgroup, median headacheday reduction was numerically higher for the intervention group (2.0 [IQR 1.04.0] vs 1.5 [IQR 1.02.25]; uncorrected P=.049), which lost statistical significance following Bonferroni adjustment (adjusted α=.0125). No statisticallysignificant symptomday reductions were observed for vestibular migraine, Ménière’s disease, or benign recurrent dizziness. The proportion of dizziness attacks accompanied by unsteadiness decreased (3.6% vs 22.0%, P=.013), and acutephase headachemedication documentation increased (50.0% vs 20.5%, P=.022). Every platformengagement metric was elevated in the intervention group: daily patient messages (+183%), medicalteam replies (+91%), patient gratitude messages (+80%). Stratified consultation volumes increased for emergency helpseeking (+643%), emotionalanxiety feedback (+325%), lifestylediet inquiries (+220%), and active headachesymptom selfreporting (+284%).

Conclusions:

Agentassisted followup improved specialtydiary adherence and objective platformmeasured patient engagement. Selected behavioral clinical outcomes improved, whereas consistent reductions in monthly symptomatic attack days were not demonstrated. Seasonrelated meteorological differences between historicalcontrol and intervention cohorts constitute an important confounding factor. Further contemporaneous prospective controlled investigations are required to verify the clinical effectiveness of this intelligentagentbased mHealth intervention. Clinical Trial: > The study has been reviewed and approved by the Chinese Clinical Trial Registry (ChiCTR) within the National Medical Research Registration and Filing Information System, with the registration number ChiCTR2600127103.