This report presents an interim analysis of a prospective multicenter observational study designed to evaluate the feasibility and utility of smartphone-based longitudinal monitoring of subjective well-being (SWI, measured using a graphic rating scale with integer values from 0 to 10) and AEs following neurosurgical procedures. The SWI is a single-item rating scale that we developed for low-burden, repeated smartphone-based assessment of overall subjective well-being. Its format is conceptually similar to that of the numeric rating scales from 0 to 10 commonly used for subjective symptoms such as pain. It was chosen to enable frequent longitudinal measurements without the burden of repeatedly completing longer questionnaires. The SWI has not undergone formal validation. The overarching aim of the complete study was to develop a statistical model quantifying the association between patient-reported well-being and relevant clinical variables, thereby characterizing the normal postoperative trajectory and how it may vary depending on clinical factors, including the type and severity of AEs. The present analysis emphasizes feasibility and descriptive statistics. Patient enrollment began in June 2023 at the Cantonal Hospital of St. Gallen (now part of HOCH Health Ostschweiz), a tertiary academic hospital in Switzerland. This interim analysis is restricted to participants enrolled at this site from June 2023 to January 2025. Additional recruitment has since commenced at 3 other major neurosurgical centers in Switzerland: University Hospital Zurich, Cantonal Hospital Graubünden, and Inselspital Bern. These results thus represent the early phase of a larger Swiss multicenter study.

Patients scheduled for one of the following four predefined elective neurosurgical procedures were screened: (1) lumbar decompression with or without sequestrectomy, discectomy, or foraminotomy (including single- or multiple-level procedures); (2) lumbar transpedicular instrumentation and fusion (with or without interbody cages, including thoracolumbar or extension to the pelvis); (3) supratentorial craniotomy; and (4) infratentorial craniotomy. Patients were eligible for inclusion if they were at least 18 years of age, able to consent, willing to provide data up to 1 year after surgery, and in possession of and capable of using a smartphone (Android or iOS); had the necessary language (German) and cognitive skills to use the OP Tracker app; and were able to provide a preoperative SWI and QoL assessment. Exclusion criteria included pregnancy, foreseeable difficulties using the OP Tracker app, presence of a condition that hindered the baseline preoperative assessment, and health conditions that rendered inclusion unsafe (eg, untreated ruptured intracranial aneurysm). Emergency surgical patients were excluded unless surgery was scheduled at least one day after admission, allowing time for proper inclusion and baseline data entry.

Following written informed consent, the OP Tracker smartphone app was installed on the participants’ devices. A study physician entered the date and type of surgery into a secure web interface, which generated a unique participant number (UPN). This UPN was used to pseudonymize the app-based data. Baseline assessments, including demographic information, surgical indication, and comorbidities, were recorded by the study team in a separate spreadsheet. Within the app, the following data were collected: date of surgery, category of surgery, SWI (rated 0-10), EQ-5D-5L questionnaire, and AEs. Each entry was time-stamped. One item of the EQ-5D-5L records self-rated QoL on a visual analog scale ranging from 0 to 100. To ensure consistency with the SWI, these values were linearly transformed to the range of 0 to 10 and incorporated as additional SWI entries. AEs were recorded using predefined categories, with an additional free-text option for events not represented by these categories, and their severity was graded by the patient using the TDN grade (Table 1) [10].

Patients received automated app reminders for data entry: daily in the first 2 weeks postoperatively (starting from the first postoperative day), every 2 days from day 15 to day 42, twice weekly up to 3 months, weekly up to 6 months, and biweekly up to 1 year after surgery. AEs could be reported at any time. If an AE was reported, the reminder frequency was reset to daily for 14 days. EQ-5D-5L reassessments were scheduled at 3 and 12 months. The full study pathway, including enrollment and follow-up schedule, is illustrated in Figure 1.

A dedicated screening log was maintained daily by the authors. To identify candidates as early as possible, a study nurse screened the neurosurgical operation schedule weekly, entering patients planned for 1 of the 4 predefined procedures into a secure study spreadsheet. This screening log was updated daily by the investigators to account for short-notice scheduling changes, ensuring that eligible patients were not missed. Recruitment took place either in the outpatient setting (at the time of surgical planning) or after hospital admission provided that surgery was not planned for the same day. For each enrolled patient, baseline demographic and clinical information (sex, age, surgical indication, and planned procedure) was recorded in a secure, pseudonymized (UPN) spreadsheet. The first data entry on the OP Tracker app was completed together with the investigators to ensure usability and correct data entry by the patient. Subsequent data entries were completed independently by the patient. Data collected via the app were automatically encrypted and uploaded to a secure Swiss-based server. Investigators accessed the data through a password-protected, web-based platform. AEs were defined as any deviation from the expected perioperative course, including intraoperative and postoperative events. They were later assessed and graded according to the TDN classification by 2 study physicians through retrospective review of medical records. A similar methodology was recently validated, with excellent inter- and intrarater reliabilities [12]. The screened medical records comprised operative reports, discharge summaries, and outpatient follow-up documentation. Laboratory and imaging data were not systematically rescreened using predefined diagnostic thresholds; AEs were recorded when explicitly documented by the treating physicians.

The primary objective of this interim study was to assess the feasibility of longitudinal, smartphone-based monitoring in a neurosurgical patient population. Feasibility was assessed descriptively based on recruitment rates, participant adherence to data entry, completeness of SWI and EQ-5D-5L data, and reliability of AE reporting via the app. As this was an exploratory feasibility assessment, no predefined quantitative thresholds for successful feasibility were specified. As part of the exploratory analysis, we assessed the suitability of future modeling approaches (mixed-effects models) to capture within- and between-patient variation in SWI values over time. Additionally, patient-reported and physician-adjudicated AEs were compared based on frequency, type, and severity. This feasibility phase did not involve formal sample size calculation. Selection bias was expected as only patients with compatible smartphones and sufficient digital literacy could participate. Recall bias was reduced through frequent notifications and display of prior SWI entries, which may introduce anchoring bias. Additionally, attrition bias is likely to occur as patients experiencing severe complications are more likely to discontinue app use, leading to underreporting of unfavorable outcomes. No external control group was used in this interim study.

All analyses were performed using R (version 4.4.0; R Foundation for Statistical Computing) [20]. Data were manipulated using tidyverse [21] and visualized using ggplot2 [22], and summary tables were formatted using gt [23]. Categorical variables are summarized as frequencies and/or percentages, and continuous variables are summarized using means or medians as appropriate. Analyses were based on available cases; missing data were not imputed. For group comparisons, the Kruskal-Wallis rank-sum test was applied to continuous variables, and the Fisher exact test was used for categorical variables. P values are reported for descriptive purposes only, and no adjustment for multiple testing was performed. A P value below .05 was considered statistically significant. Using nlme [24], a mixed-effects model was constructed to account for differences between patients and the correlation of repeated measurements over time. The influence of time since surgery, age, sex, type of surgery, and both patient-reported and physician-reported AEs on well-being scores was examined. The model included random intercepts and random slopes for time (months) for each patient and accounted for the fact that repeated measurements from the same patient tend to be correlated using an AR1 structure.

This study was approved by the relevant regional ethics committee (Business Administration System for Ethics Committees [BASEC] 2022-01510; Ethikkommission Ostschweiz [EKOS] 22/131); registered at ClinicalTrials.gov (NCT06352710); and conducted in accordance with the Declaration of Helsinki, the principles of good clinical practice, and the Swiss Human Research Act and Human Research Ordinance. No study-specific clinical interventions were performed, and standard care pathways remained unchanged. All participants gave written informed consent. Additionally, this study complied with the mERA (mobile health evidence reporting and assessment) checklist (Checklist 1) to ensure thorough reporting of digital health research and with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) statement for observational studies (Checklist 2) [25,26].

Between June 12, 2023, and January 31, 2025, a total of 626 patients scheduled for neurosurgical procedures at the Cantonal Hospital of St. Gallen were screened for eligibility. Of these 626 patients, 100 (16%) met all inclusion criteria and were included. Reasons for exclusion were preoperative inclusion not feasible due to logistical constraints (eg, same-day surgery; 298/626, 47.6%), anticipated difficulties using the app (85/626, 13.6%), lack of access to a compatible smartphone (60/626, 9.6%), patient refusal to participate (49/626, 7.8%), age below 18 years (7/626, 1.1%), incapacity to provide informed consent (6/626, 1%), current pregnancy (4/626, 0.6%), or unspecified reasons (17/626, 2.7%).

The inclusion rate peaked in the third month after study initiation, followed by a decline and then a relatively stable rate over the subsequent months (Figure 2). The drop coincided with an increasing number of patients admitted on the day of surgery, making inclusion impossible due to the requirement for app installation and baseline data entry prior to surgery.

Of the 100 included patients, 68 (68%) underwent spinal surgery, and 32 (32%) underwent cranial surgery. The most common procedures were lumbar fusion (35/100, 35%), followed by lumbar decompression (33/100, 33%). There were 32 craniotomies, of which 28 (87.5%) were supratentorial. The median age at surgery was 64.0 (IQR 52.95-71.6) years, and 45% (45/100) of the participants were female. No significant age difference was observed between subgroups. More female participants underwent spinal fusion, whereas more male participants underwent lumbar decompression or craniotomy. Patients undergoing craniotomy had a markedly higher preoperative QoL compared with those undergoing spinal surgery. These results are summarized in Table 2.

Postoperative data entry through the OP Tracker app was achieved by 86% (86/100) of the participants. The median follow-up period reached 3.2 (IQR 0.2-11.3) months, during which patients submitted a median of 38 (IQR 4-65) SWI entries. The completeness of SWI data was high in the early postoperative period, with 84% (84/100) of scheduled entries completed within the first month following surgery. This declined moderately over time, with completion rates of 59% (59/100) at 3 months, 51.5% (50/97) at 6 months, and 43.7% (31/71) at 12 months. A total of 42% (42/100) of the participants provided more than 50 SWI entries, whereas 35% (35/100) provided less than 10 (Table 3). EQ-5D-5L QoL data were collected at 3 time points. Preoperative responses were obtained from 85% (85/100) of participants, responses at 3 months were obtained from 43% (43/100) of patients, and responses at 12 months were obtained from 32.4% (23/71) of patients. Individual postoperative SWI trajectories for all patients with at least 5 entries are shown in Multimedia Appendix 1.

A total of 30 AEs were self-reported via the OP Tracker app by 15% (15/100) of the participants. Because patients could enter free-text descriptions, similar complications were sometimes described differently, and in some cases, the same AE was reported multiple times. To avoid overestimation, we grouped similar descriptions and counted repeated reports only once per patient, resulting in 22 unique AEs from 15 categories (Table 4).

For comparison, hospital records of all included patients were reviewed by 2 investigators. In total, 44 AEs were identified in 36% (36/100) of the patients, and 9 (20.5%) of these events were also correctly reported by patients through the app. In these cases, patients assigned the correct TDN grade in 66.7% (6/9) of cases. The median TDN grade was 2.0 for both physician- and patient-reported AEs (IQR 1.75-3.0 and 2.0-3.0, respectively), with no differences across surgical subgroups. Most investigator-reported AEs were mild (TDN 1 or 2; 30/44, 68.2%), and there were no grade 4 or 5 events. Patients and investigators tended to report different types of AEs: the most frequent patient-reported events were sensory or motor deficits (n=4), intraoperative dural tears (n=3), urinary tract infections (n=2), and persistent or aggravated pain (n=2), whereas the most frequent investigator-reported events were anemia (n=6), pseudarthrosis or implant-related complications (n=6, consisting of 1 cage displacement, 1 nerve root compression by bone graft material, 2 pseudarthrosis, and 2 screw loosenings), sensory or motor deficits (n=4), and intraoperative dural tears (n=4). The longitudinal relationships between patient well-being and AEs are illustrated in Figure 3.

When stratified by surgery type (Table 5), patients who underwent lumbar fusion had the highest rate of investigator-reported AEs (29/35, 82.9%), whereas those who underwent lumbar decompression had the lowest (6/33, 18.2%). Interestingly, lumbar fusion was also the group with the lowest proportion of patient-reported AEs (4/35, 11.4%).

A mixed-effects model was computed to evaluate factors influencing postoperative well-being (Figure 4). As expected, time since surgery was positively associated with subjective well-being, reflecting gradual recovery during follow-up. No other factors reached statistical significance as their 95% CIs crossed 0. However, there was a tendency toward lower well-being among female patients, those who underwent spinal fusion, and those who experienced AEs, whereas patients who underwent craniotomy tended to report higher well-being. Importantly, age had no measurable impact on postoperative well-being, suggesting that perceived recovery was not age dependent in this population.

This interim analysis demonstrates the feasibility of smartphone-based longitudinal monitoring of postoperative well-being and AEs in an elective neurosurgical cohort while also identifying challenges related to recruitment and sustained long-term participation. The 100 enrolled patients submitted 4354 QoL entries (mean of almost 50 per patient). Data completeness was highest during the early postoperative period and decreased over time. Patients reported 22 unique AEs compared with 44 identified by investigators, with limited overlap. Most AEs were mild (TDN grade 1-2; 30/44, 68.2%), and none were life-threatening or highly disabling. A preliminary mixed-effects model identified time as the only positive predictor of SWI scores. Together, these findings support the feasibility of patient-driven digital monitoring, demonstrate its potential to provide richer insights than conventional follow-up by complementing physician documentation with the patient perspective, and underscore the importance of sustaining engagement during longer-term follow-up.

Postoperative outcome research and complication monitoring continue to face major methodological challenges. First, traditional methods of tracking postoperative outcomes are static and resource intensive (eg, questionnaires during rare follow-up consultations at fixed time points after surgery), leading to sparse data, vulnerability to missing information, and an incomplete understanding of the normal postoperative course. Second, AE reporting and monitoring are hindered by the lack of detection tools as they are usually defined as deviations from the normal postoperative course, which, as stated above, remains unknown. Third, the severity of AEs is usually reported regardless of the disability experienced by patients. The introduction of the TDN classification addressed this gap [10], but the influence of AEs on QoL has not yet been prospectively evaluated. Together, these limitations greatly hinder meaningful outcome assessment, informed patient counseling, and data-driven improvement of medical care.

Addressing these challenges requires new data collection methods that capture patients’ well-being in their own environment and detect deterioration early. In this context, the near ubiquity of smartphones offers a practical platform for such continuous monitoring. Digital health tracking devices, including wearables and smartphone-based sensors, have been shown to detect complications earlier and improve postoperative care in general surgery [16,27] and are increasingly applied in neurosurgical patients as well [28]. Still, no large-scale, patient-driven mobile postoperative monitoring solution has been implemented to our knowledge.

Our solution is an easy-to-use smartphone app that enables close monitoring of life quality fully integrated into patients’ environment. The early findings from our single-center experience offer several insights that may help shape future integration of digital health solutions in the clinical setting. The app enabled us to gather thousands of life quality data points, which would be completely impossible in the usual clinical routine. The successful enrollment and retention of patients validate the practicality of integrating a smartphone app into the care pathway, and the median age of participants (64.0, IQR 52.95-71.6 years) compared adequately with that of a large neurosurgical series from Switzerland [29]. Our results show that many neurosurgical patients are both able and willing to engage in digital health tracking, highlighting the added value of combining patient-reported and clinical data. Engagement was more sustained for the frequently prompted SWI than for the EQ-5D-5L, which was administered only at predefined follow-up time points. Missed EQ-5D-5L prompts were therefore less likely to be recovered through subsequent notifications, suggesting that repeated reminders for infrequently administered outcome measures may improve long-term completeness. Patients and physicians did not always report the same AEs: clinicians often noted laboratory or radiographic findings (eg, anemia, acute kidney injury, and pseudarthrosis) that patients did not perceive as complications, whereas patients mainly reported symptomatic or functionally relevant issues such as neurological deficits, pain, or bed rest after a dural tear. This divergence illustrates that patient-reported outcomes focus on the experienced and functional burden of surgery, complementing physician-based reporting. Using the TDN grade helped bridge these perspectives by prompting patients to consider new deficits and disabilities rather than medical interventions alone. Both patients and physicians predominantly reported low-grade (TDN 1-2) events, reflecting the predominance of minor complications in this elective cohort, and the 20.5% (9/44) overlap indicates that relying on either source alone would miss a substantial part of the postoperative picture. Together, these findings support a hybrid model of outcome monitoring that integrates patient-driven digital reporting with traditional clinical follow-up to achieve a more comprehensive understanding of recovery.

This single-center analysis represents the initial phase of an ongoing multicenter Swiss study aimed at characterizing the postoperative course by analyzing the relationship between patient-reported QoL and clinical variables such as type of surgery and severity of AEs. Subsequent iterations of the OP Tracker app could incorporate configurable procedure- and diagnosis-specific outcome measures. Building on this foundation, future integration of patient-driven self-reporting with passive data streams from wearables, biosensors, and electronic health records could enable real-time, adaptive feedback loops between patients and care teams. Such connected, multimodal approaches have the potential to redefine postoperative care by enabling continuous, personalized recovery monitoring and early detection of complications at scale.

This analysis has several limitations inherent to its design and scope. It included the first 100 patients from a single center, with a predominance of spinal procedures, which may limit generalizability. Because participation required the use of a smartphone, selection bias is likely. Patients with limited digital access or literacy, significant neurological impairment, or poor postoperative recovery may therefore be underrepresented. This is particularly relevant because these patients may also be at increased risk of postoperative complications, limiting the generalizability of our findings to the broader neurosurgical population. In addition, severe or higher-grade AEs are expected to be underrepresented in patient-reported data, particularly when the event itself limits the patient’s ability to interact with the app. This highlights that smartphone-based self-reporting should complement rather than replace clinical surveillance. Similarly, AEs that were asymptomatic or perceived as expected may have been underreported by patients. This is reflected in the lower number of patient-reported AEs compared with physician documentation. However, the parallel physician assessment of AEs mitigates this limitation by providing a more complete overview and showing a similar TDN grade distribution, suggesting that patients can use the system meaningfully for less severe complications. The inclusion of heterogeneous spinal and cranial procedures was appropriate for assessing the feasibility of a broadly applicable monitoring platform, but differences in baseline characteristics, recovery trajectories, and AE profiles limit the interpretation of pooled analyses and the exploratory mixed-effects model. Procedure-specific analyses will require larger subgroup sample sizes. Finally, as an interim single-center analysis, these results should be interpreted as preliminary but supportive of the feasibility and potential of patient-driven digital monitoring.

In this interim analysis, smartphone-based postoperative monitoring was feasible and engaging for neurosurgical patients, generating high-density, patient-centered data that complement traditional clinical follow-up. The integration of patient-reported well-being and AE tracking offers a richer, more nuanced view of recovery and highlights the potential of digital tools to close long-standing gaps in outcome research. As the multicenter study progresses, this approach may pave the way toward data-driven, patient-centered systems for continuous and personalized monitoring of postoperative recovery, although its impact on clinical decision-making and patient outcomes remains to be established.