This cross-sectional study used a nonprobability convenience sampling method. Participants were recruited through online platforms (eg, Facebook, LINE) and professional networks, including outreach by health care professionals.

Participants were eligible if they were adults aged 20 years or older, resided in Taiwan, had experience using SMS-capable mobile phones, and completed the online questionnaire with informed consent. Individuals were excluded if they lacked SMS access, had known cognitive impairments, or reported working in health care–related professions (eg, hospitals or clinics) to minimize potential response bias. The structured questionnaire was distributed electronically, and the data were collected anonymously over a 14-day period from February 25 to March 10, 2025. A total of 1152 responses were received, of which 1095 were valid, yielding a valid response rate of 95.05%.

Descriptive statistics summarized participant characteristics. Chi-square tests and one-way ANOVA were used to examine bivariate associations between participant characteristics and SMS reminder preferences. To control for inflated type 1 error due to multiple comparisons, the Benjamini-Hochberg false discovery rate (FDR) procedure was applied to the bivariate tests [24].

To adjust for potential confounding and identify independent predictors of SMS reminder preference, we fitted a multinomial logistic regression model with SMS reminder preference as the outcome. The general reminder SMS was specified as the reference category. Demographic characteristics and health care utilization variables were entered simultaneously as covariates, and the results are reported as adjusted odds ratios with 95% CI. Statistical significance was set at P<.05 (2-tailed). All analyses were conducted using SPSS (version 29.0.1.0).

This study was reviewed and approved by the Institutional Review Board of Kaohsiung Medical University Chung-Ho Memorial Hospital (IRB number: KMUHIRB-E(I)-20240047; approval date: February 3, 2025). Electronic informed consent was obtained from all participants prior to survey submission. Participation was voluntary, anonymity was ensured, and no identifiable personal information was collected. All study procedures adhered to institutional ethical guidelines and the principles of the Declaration of Helsinki.

As shown in Table 3, nearly half (n=520, 47.5%) of the respondents preferred reminders referencing “Prior MAs” behavior, followed by general reminders (n=341, 31.1%). Fewer participants selected messages with an empathetic tone, including provider-focused empathy (n=48, 4.4%) and patient-focused empathy (n=60, 5.5%). Relational reminders, such as physician follow-up (n=53, 4.8%) and supportive reminders (n=73, 6.7%), accounted for a combined total of 11.5% of the preferences. These results indicate a predominant preference for practical and behavior-linked reminder formats over affective or relational messaging.

Because 4 SMS categories (provider-focused empathy, patient-focused empathy, physician follow-up, and supportive reminders) were selected by relatively few participants, the corresponding subgroup cell sizes in the chi-square analyses were small. Therefore, statistical comparisons involving these categories should be interpreted with caution.

Male respondents showed relatively higher proportions selecting supportive or relational message types, although these differences were not statistically significant after FDR correction.

Analyses of demographic and health care utilization characteristics showed several initial associations with SMS reminder type preferences. As summarized in the cross-tabulations, gender, age group, education level, occupation, residential region, and living arrangement each demonstrated significant differences across message categories in unadjusted chi-square and ANOVA analyses (Table 4). For example, women and younger adults more frequently selected general or prior missed appointment reminders, whereas older adults showed comparatively higher proportions selecting empathetic or supportive formats. Occupational variation was also observed, with clerical and professional workers tending toward standard reminders, while senior executives reported a greater share of emotionally supportive preferences. These subgroup differences, however, did not persist following correction for multiple comparisons.

Health care utilization variables exhibited similar patterns. Participants with 1 to 3 outpatient visits in the past 3 months tended to prefer neutral, task-oriented reminders, while those with at least 1 recent MA were more likely to choose messages explicitly referencing prior absences (Table 5).

To address the potential inflation of type 1 error due to multiple comparisons, the Benjamini-Hochberg FDR correction was applied across all demographic and utilization variables. After adjustment, only age and gender remained statistically significant predictors of SMS preference. All other variables, including education, occupation, region, living arrangement, outpatient visit frequency, and missed appointment history, no longer met the adjusted significance threshold. Multinomial logistic regression yielded consistent results, confirming age and sex as the only independent predictors after controlling for all covariates.

Although initial chi-square and ANOVA results suggested that several demographic and health care utilization variables were associated with SMS reminder preferences, these patterns did not remain statistically stable after controlling for multiple comparisons. After applying the Benjamini-Hochberg FDR correction, only age and sex remained significant predictors, whereas education level, occupation, residential region, living arrangement, outpatient visit frequency, and missed appointment history were no longer statistically significant (Table 6).

To identify independent predictors of SMS reminder preference while simultaneously adjusting for all covariates, a multinomial logistic regression model was constructed. The full results of this multivariable analysis are presented in Table 7, including adjusted odds ratios (aORs), 95% CI, and P values.

As shown in Table 7, the multinomial logistic regression analysis identified age as the only statistically significant independent predictor of SMS reminder preference after simultaneous adjustment for all covariates. Participants younger than 50 years were significantly more likely to prefer alternative reminder message types compared with the general reminder SMS (aOR 1.64, 95% CI 1.18‐2.28; P=.003). In contrast, sex did not remain independently associated with reminder preference after multivariable adjustment (aOR 1.23, 95% CI 0.87‐1.73; P=.24). Other sociodemographic characteristics and health care utilization variables, including education level, employment status, residential region, living arrangement, outpatient visit frequency, and recent MAs history, were not statistically significant in the fully adjusted model.

The primary objective of this study was to examine public preferences regarding SMS reminder content in preparation for a forthcoming large-scale randomized controlled trial. In the planned randomized controlled trial, participants will be randomly allocated to either an intervention group, which will receive a standardized SMS reminder seven days prior to their scheduled appointment containing details such as the appointment date, time, department, and queue number, or a control group that receives no reminder. Insights from this study will inform the development of effective SMS content strategies to enhance outpatient appointment adherence. The following sections address several key issues that merit further discussion.

The observed preference for concise, behaviorally framed SMS reminders, particularly the general reminder and prior MA formats, was most prominent among women, younger adults, and highly educated respondents. These patterns align with the demographic profile of the study sample, in which the majority were female (n=851, 77.7%), under age 50 (n=720, 65.8%), and held a college degree or above (91.1%).

Prior literature indicates that women are more likely than men to assume caregiving responsibilities within households, maintain greater engagement with preventive health services, and demonstrate higher willingness to participate in self-administered online questionnaires related to health behaviors and health care use [25-27]. These factors may contribute to their increased representation in digital health surveys.

Furthermore, women consistently report higher health literacy and mobile engagement for health-related purposes, including appointment scheduling, medication reminders, and symptom monitoring, compared with men of the same age groups [28]. Studies have also shown that women exhibit stronger preferences for receiving health information via SMS or other asynchronous communication channels, which may further elevate their likelihood of responding to an online survey focused on SMS reminder content. In contrast, men tend to participate at lower rates in health communication surveys unless participation is tied to an active medical condition or clinical follow-up requirement [29].

Collectively, the predominance of women in this sample likely reflects well-documented patterns in digital health engagement rather than a study-specific recruitment artifact. Nevertheless, the skewed sex distribution requires the cautious interpretation of aggregated preference patterns. To address this concern, sex was explicitly incorporated as a key covariate in the multinomial logistic regression analysis. After multivariable adjustment, sex did not remain independently associated with SMS reminder preference, indicating that the observed preference patterns were not solely driven by sample composition.

A minor discrepancy was observed between outpatient visit reports and MA data. Specifically, 17.5% (n=192) of the respondents reported no outpatient visits in the past 3 months, but 15.7% (n=172) reported at least 1 MA. This inconsistency may reflect recall bias or variable interpretations of MAs, such as canceled or forgotten visits not formally captured by hospital systems. Our findings echo those of de Reuver and Bouwman [30], who observed that type 1 and type 2 reporting errors are common in self-reported utilization data.

The chi-square analyses indicated that participants most frequently selected general reminders and prior MA messages, while empathy-oriented or relationally framed messages were selected far less often. This suggests that patients prioritize clarity and actionable information over affective framing when receiving appointment-related communication. These findings are consistent with Hallsworth et al [16], who demonstrated that reminders containing cost-related information and clear behavioral cues were more effective than emotionally framed messages.

The results demonstrated meaningful variation in SMS reminder preferences by age and sex, whereas other demographic and behavioral characteristics did not show stable associations after correction for multiple testing.

After correction for multiple testing, age and sex remained statistically significant in FDR-adjusted univariate analyses. However, when all covariates were entered simultaneously into the multinomial logistic regression model, age emerged as the only independent predictor of SMS reminder preference. Sex did not retain statistical significance after multivariable adjustment.

After multivariable adjustment, age emerged as the only independent predictor of SMS reminder preference, whereas sex did not retain statistical significance. Other variables, including education, occupation, residential region, living arrangement, outpatient visit frequency, and recent MAs, lost statistical significance following the FDR adjustment.

Descriptively, men demonstrated a more even distribution of preferences across message types; however, these differences did not remain statistically significant after multivariable adjustment and should be interpreted cautiously.

These findings indicate that designing personalized SMS strategies based primarily on age and sex may be more appropriate than tailoring messages by education level or utilization history. This pattern is consistent with Crutchfield et al [31] and Mohammed Selim et al [32], who observed that sociodemographic characteristics influence responsiveness to reminder messages. An Israeli randomized controlled trial further found that older adults and those with prior MAs were more responsive to emotionally framed messages [33]. Junod Perron et al [34], however, emphasized that communication mode (SMS vs telephone) may influence engagement differently than message framing.

Because multiple chi-square tests were conducted, the Benjamini-Hochberg FDR procedure was applied to reduce the risk of type 1 error. After correction, only age and sex remained statistically significant predictors of SMS preference. Residential region, living arrangement, and all health care utilization variables, including outpatient visit frequency and recent MAs, were not significant after correction.

These results were corroborated by the multinomial logistic regression model, in which age and sex remained the only independent predictors after mutual adjustment. Overall, these findings indicate that although unadjusted analyses suggested broader subgroup differences, only age- and sex-based differences demonstrated sufficient stability to warrant consideration in tailored SMS design.

It should be noted that several message types, particularly empathy-based formats, had limited subgroup counts. The observed associations for these categories may therefore lack statistical robustness and should be interpreted conservatively. The revised text avoids overstating preference differences in low-frequency groups.

To ensure the appropriateness and clarity of SMS content prior to trial implementation, this study used item-level content validity indices based on 9 communication design principles derived from existing literature [17,19,23]. A multidisciplinary expert panel of 12 evaluators assessed the messages, exceeding recommended minimum panel sizes for content validation.

Although the panel did not include formal patient representatives, all members had extensive experience interacting with patients in outpatient or hospital settings. These perspectives provided practical insights grounded in real-world communication processes. This limitation is acknowledged, yet expert-based content validation remains standard practice in early formative research.

The decision to dichotomize age at 50 years was justified by both statistical and contextual considerations. Statistically, this threshold yielded significant differences in SMS preferences. Contextually, adults around age 50 years in Taiwan frequently assume dual caregiving responsibilities for both aging parents and children, which may influence their engagement with health care communication [35].

This study relied on convenience sampling through an online survey, which may limit representativeness. The predominance of female respondents reflects a well-documented pattern in online and digital health surveys, in which women are more likely than men to participate, particularly in studies related to health behaviors and health care communication [36,37]. To address this imbalance, sex was explicitly incorporated as a covariate in both univariate analyses and the multinomial logistic regression model. After multivariable adjustment, sex did not remain independently associated with SMS reminder preference, suggesting that the observed patterns were not solely driven by sample composition. Nevertheless, absolute preference proportions should be interpreted with caution, particularly for male patients. Future studies may enhance representativeness through targeted recruitment strategies and the inclusion of more inclusive gender response options, in line with emerging best practices in survey methodology [38].

This study demonstrates that preferences for SMS appointment reminder content vary systematically among outpatients, with age emerging as the most robust independent predictor. General reminders and prior missed appointment messages were most consistently preferred, whereas empathy-based or relational formats were selected far less frequently. After applying FDA correction and conducting multinomial logistic regression, age remained the only stable independent predictor of SMS reminder preference, whereas sex did not retain statistical significance after multivariable adjustment. Beyond predictor effects, the observed preference for concise, action-oriented messages underscores the potential importance of clarity and behavioral specificity when designing SMS reminders in Taiwan’s universal health coverage context. Content validity assessment further confirmed that the 6 SMS prototypes possess adequate semantic clarity and cultural appropriateness for use in subsequent interventional research.

Although convenience sampling limits generalizability, the results offer valuable formative evidence to guide the development of SMS interventions aimed at improving appointment adherence. Future randomized controlled trials should evaluate the real-world effectiveness of these demographically tailored and behaviorally informed reminder formats in reducing outpatient no-show rates and supporting more efficient use of ambulatory care resources.