We followed the medical research council’s framework for evaluating complex interventions [12] and prospectively registered our study at clinicaltrials.gov (NCT05564195) and published our study protocol [13]. We report in accordance with reporting guidelines for nonrandomized feasibility studies [14] and relevant items from the CONSORT (Consolidated Standards of Reporting Trials) 2010 statement extension for pilot and feasibility trials [15].

We conducted a nonrandomized single-group feasibility study with the purpose of assessing key components for a future, larger cohort study [14]. The study was carried out from January 2023 to November 2024 in Stockholm County, Sweden. The hospital where the patients were recruited specializes in abdominal surgery and handles approximately 4700 outpatient visits annually.

Registered nurses at the surgery clinic and research unit screened and identified potential participants who met the inclusion criteria at the preoperative visit, provided study information, and obtained written informed consent. Inclusion criteria were patients ≥60 years undergoing planned surgery lasting a minimum of 60 minutes. Exclusion criteria included Mini-Mental State Examination (MMSE) score <23, current psychiatric disease (eg, bipolar disorder), central nervous system disease (eg, stroke), uncorrected hearing or visual impairments, motor impairment in the dominant hand, defective color vision, and lack of proficiency in the Swedish language. A target sample size of 50 participants was planned, based on a recommended minimum of 30 participants for pilot studies to provide initial estimates of outcome variability and feasibility for future trials [16]. A Swedish version of the MMSE was administered at the beginning of the study before obtaining written permission from Psychological Assessment Resources (PAR). This administrative oversight was later corrected, and the required authorization has since been secured. The MMSE is a copyrighted instrument and may not be reproduced or used in any form or language without prior written permission from PAR.

Quantitative data investigated recruitment and retention process, usability, cognitive performance, including clinician and patient-reported outcome measures. We collected data at baseline (T0) and postoperatively including follow-ups at the hospital 1‐3 days (T1), 3‐5 weeks (T2), and at 6 months (T3) at the patient’s home or at the premises of Karolinska Institutet. All the outcome measurements included are presented in Table 1. As some instruments lacked validation in older adults in a perioperative context, their use in this study was exploratory.

Qualitative data explored the acceptability, usability, and perceived value of the digital cognitive test battery by collecting data from 2 sources:

The first author (AA) conducted the interviews with the patients, and the last author (LB) conducted the interviews with the nurses. We developed and followed semistructured interview guides (Multimedia Appendix 1) for all interviews, which contained open-ended questions and suggestions for additional probing questions. The interview guide for the nurses was based on Sekhon et al’s [27] framework for acceptability, which includes 7 constructs: affective attitudes, burden, ethical aspects, intervention coherence, opportunity costs, perceived effectiveness, and self-efficacy. After the initial interviews, the guide was discussed and evaluated within the research group. Each interview was audio-recorded and transcribed verbatim by a transcription company.

Mindmore offers different cognitive test batteries depending on the patient group and procedure [7]. The tests included in this study are listed in Table 2. Each test begins with an introduction, followed by a practice trial with feedback and then the actual test trial. The 4 tests take approximately 15 minutes to complete and are conducted on a 10.1-inch Windows touchscreen study-specific tablet. The capacitive touchscreen records detailed information such as timing, pauses, and finger lifts from the screen. Further, speech recognition is used for screening verbal memory. The patients underwent the cognitive test battery at T0, T1, T2, and T3.

An overview of the feasibility outcomes and progression criteria [28] is presented in Table 3. Our progression criteria (eg, warranting a recommendation to proceed to a larger trial) were based on our previous research with patients undergoing total hip arthroplasty [29] and previous studies including patients undergoing colorectal surgery [30,31]; consequently, we set the following progression criteria: recruitment ≥20% of available patients, dropouts ≤30%, and usability ≥75 on the System Usability Scale.

Descriptive statistics included numbers and percentages for categorical variables, means, median, SD, for continuous variables, and 95% CIs where appropriate. Analysis of normal distribution was performed with Q-Q plots, histograms, and the Shapiro-Wilk test. We applied the nonparametric Friedman test on repeatedly measured data. We defined cutoffs for dNCR/p-NCD following the recommendations by Borchers et al [6]: mild decline at 1 to 2 SD and major decline >2 SD in at least 2 test parameters at postoperative time points 3‐5 weeks or 6 months. We considered a 2-sided P value of <.05 as statistically significant. All statistical analyses were performed in IBM SPSS version 28 (IBM Corp).

Three authors with experience in qualitative methods (AA, MS, and LB) analyzed the qualitative data through an iterative process. We conducted a deductive content analysis as described by Kyngäs and Kaakinen [32] by applying the theoretical framework of acceptability [27] on our data. Initially, we created an analysis matrix based on the 7 constructs of acceptability and added a dimension for study design, feasibility, and recruitment. Authors AA, MS, and LB thoroughly read each transcribed interview multiple times to familiarize themselves with the data and assess the depth of information. LB organized the data using NVivo (version 15, QSR International Pty Ltd), and coded quotes related to our research questions using the analysis matrix. AA validated the codes independently in NVivo. Subsequent discussions with MS ensured coding decisions and researcher triangulation. MS created categories inductively according to the constructs in our matrix, with AA and LB reviewing and refining them collaboratively. AA translated the final meaning units from Swedish to English, and LB selected illustrative quotes for inclusion. Discussion of initial categories and reflections with the co-authors of the manuscript led to the re-organization of our meaning units and categories within the matrix. All authors’ diverse clinical and research backgrounds, including anesthesia, surgery, postoperative care, intensive care, and neuropsychology, provided complementary perspectives during analysis. The content analysis is presented in Table 4.

We obtained an ethical permit (2022-03593-01) from the Swedish Ethical Review Authority before initiating the study. All procedures complied with the World Medical Association’s Declaration of Helsinki. Written informed consent was obtained from all participants before inclusion. Data were pseudonymized to ensure confidentiality. Participation was voluntary, without compensation, and participants could withdraw at any time.

Twenty-four patients were included in the study: 13 women and 11 men. The mean age was 77 (SD 6.5) years. Half of the patients (12/24, 50%) were classified as “very fit,” and none were classified as frail according to the Clinical Frailty Scale (CFS). Most of the patients had obtained a university degree and had an advanced occupational background: two patients were still working at the time of data collection. Most patients (20/24, 83%) underwent laparoscopic colorectal surgery for cancer under general anesthesia combined with spinal anesthesia. Eight patients developed postoperative complications, such as wound infections. Two of these patients required resurgery, and one of them also needed intensive care. At T2, 3 patients were treated with or had recently completed chemotherapy, one patient had developed an autoimmune disease, and 2 had a recent infection. Patient characteristics are presented in Table 5. The 5 registered nurses interviewed were between 35 years and 60 years old, with 12-29 years of experience in their profession.

The set progression criteria were met and the digital cognitive test battery was deemed feasible; System Usability Scale Score ≥75, recruiting ≥20% of available patients and dropouts <30%.

Recruitment took place over a period of 1.5 years, from January 2023 to June 2024. Data collection concluded in November 2024. Out of 117 patients screened for eligibility (Figure 1), 20.5% (n=24) were included in the study, dropout rates were 12% at T2 and 27% at T3.

Twenty patients completed the SUS at T2, the mean score was 87 (SD 17.9; 95% CI 78.9‐95.2), indicating excellent usability [33]. The qualitative results are presented below.

Three patients fulfilled the criteria of POD for only one day while in the surgical ward: 2 on the first postoperative day and one on the third postoperative day. No further cases of POD were observed during the subsequent postoperative days in the hospital. There were statistically significant changes in cognitive test results between the preoperative and postoperative follow-ups in the attention and tempo domain and the executive functions domain. The patients’ z scores are presented in Multimedia Appendix 2. No patients met the criteria for dNCR or p-NCD at the postoperative follow-ups. Cognitive raw scores from the different data points are presented in Multimedia Appendix 3.

Preoperatively, one patient fulfilled cut-off criteria for depression, 4 patients at T2, and one at T3. However, on a group level, the GDS scoring was low overall. Functional status scores with World Health Organization Disability Assessment Schedule (WHODAS) 12 2.0 were highest at T2. The patients’ highest scoring symptoms on the Swedish Quality of Recovery questionnaire (SwQoR) were in the cognitive and emotional domain eg, worry, concentration, and well-being. There were no statistically significant changes between preoperative and postoperative measurements of depression, functional capacity, and quality of recovery (Multimedia Appendix 2).

This was a nonrandomized feasibility study aimed at assessing a digital, self-administered cognitive test battery in older adults undergoing abdominal surgery. Recruitment and retention proved challenging, with rates falling below expectations. The 1.5-year recruitment period turned out to be unfeasible, and a single-center approach is not viable for a larger-scale trial. One contributing factor to the shortfall in sample size was that 58 eligible patients were not approached for participation, in part because some did not attend a preoperative in-person consultation before their scheduled surgery. Similar recruitment challenges have been reported in feasibility studies involving patients with cancer diagnoses [10,34,35].

For a future trial, a multicenter recruitment model using both clinical and research staff may be necessary [36]. Consent to participate can be performed remotely via an invitation letter or email to the patient before the preoperative consultation [36], and eligibility screening could be performed remotely using validated tools, such as the telephone-administered Montreal Cognitive Assessment [37,38] instead of in-person MMSE. Online advertisement such as Google Search Engine may further expand reach [39].

Other contributing factors behind low recruitment rates were staff-related: nurses refrained from inviting some patients to participate. Their decisions were influenced by their perceptions of patient receptiveness, emotional considerations due to their cancer diagnosis which aligns with previous research in patients with lung cancer diagnoses [40], and patients’ familiarity with digital devices. Additional barriers included logistical disruptions, such as hospital relocation.

The participants highlighted barriers and facilitators to implementing cognitive assessments in the perioperative setting. Some nurses expressed uncertainty about how to respond when a patient scored low on the cognitive tests, emphasizing the need for structured guidance and staff training. This aligns with findings from a study on cognitive assessments in primary care [41]. Ethical and implementation challenges can be addressed through staff education and decision-support interventions; previous evidence suggests these approaches have shown mixed success in improving management of cognitive disorders [42]. Moreover, factual knowledge alone may not be sufficient; addressing clinicians’s attitudes and implicit biases has also been shown to influence the management of cognitive disorders [43].

In terms of acceptability and usability, the digital cognitive test battery was considered easy to use and comprehend, with a high perceived value from both patients and nurses. Patients rated the digital cognitive test battery as having excellent usability, which aligns with previous findings [7]. Although cognitive screening is not performed routinely in older adults undergoing surgery, it provides actionable information to identify at-risk patients [44]. Cognitive test results indicating patients are at risk can guide preventative strategies, including implementation of the Hospital Elder Life Program for postoperative delirium [45], referral to geriatric specialists, or integration of a multidisciplinary perioperative care team [44].

No patient developed dNCR or p-NCD following surgery. The subtest assessing attention and processing speed improved over time; however, this finding must be interpreted with caution. Given the repeated exposure to the same tests, practice effects may contribute to observed improvement [46]. The small sample size and lack of a control group in our study constrain interpretation. Future trials should consider alternate versions of similar tests or control groups. Several patients had a high educational and occupational background, which reflects a cognitively robust sample. Patients with low cognitive reserves are at higher risk of developing postoperative cognitive disorders compared to those with high cognitive reserves [47,48].

This study had some strengths. To the best of our knowledge, this was the first time a digital, self-administered cognitive test battery was tested for feasibility in this clinical context. By deciding a priori progression criteria, we reduced the risk of designing an ill-informed larger trial in the future [28].

We acknowledge the limitations in this study. The cohort was relatively homogeneous and cognitively robust. This homogeneity limits the external validity, as feasibility in this group may not reflect outcomes in frailer or more diverse perioperative populations, particularly those with lower cognitive reserves. Moreover, the high usability scores should be interpreted cautiously, as they may not translate to patients with limited digital literacy. Additional limitations include the absence of a control group, which restricts interpretation of cognitive outcomes, as improvements in attention may partly reflect practice effects rather than true changes. Moreover, some of the clinician- and patient-reported outcome measurements used in this study have not been validated in this population and setting. Recruitment difficulties and the modest sample size posed additional limitations, and we acknowledge that recruitment was, in many cases, not feasible. We have a few inconsistencies with the published study protocol [13] before starting our study, as we extended the data collection and decided to include nurses to assess the acceptability and usability of the digital cognitive test battery.

The digital, self-administered cognitive test battery was found to be feasible, acceptable, and usable in older adults undergoing abdominal surgery. However, recruitment challenges and the homogeneous sample limit the generalizability of these findings. Implementation in clinical practice would require structured staff training and ethical considerations. Future trials should include larger, more diverse samples, consider control groups or alternate test versions to account for practice effects in the cognitive tests.