Determining the impact of postoperative complications in neurosurgery based on simulated longitudinal smartphone app-based assessment

This feasibility study was designed as a prospective, monocentric observational study. Participants were healthy individuals without prior disease or injury of the brain or spine. Since this study enrolled no patients, no approval by an institutional review board was required. We planned to recruit 15 subjects for this analysis.

We recruited volunteers by contacting coworkers, friends, and family members. Most of them either had a professional background in health care or in software development. The participants received a handout, containing general information about the app and the study aims, an installation manual, and a user guide.

To be included, subjects had to fulfill the following criteria:

Prior diseases or injuries of the brain or spine were our exclusion criteria.

For each participant, we mapped out a scenario, defining the type and date of surgery and describing the postoperative outcome. Subjects were required to simulate that outcome as realistically as they could. We equally distributed favorable, ordinary, and unfavorable postoperative outcomes among the participants. Those simulating an unfavorable outcome were the only participants required to enter postoperative complications. For each entered complication, we specified a realistic description, timeframe (days of onset and duration), and impact on the subjective well-being. We detailed the complication’s impact on the subjective well-being with descriptions concerning health-related quality of life and activities of daily living.

At the end of the runtime (12 weeks postoperative), we asked the participants to fill out a questionnaire, using Google Forms® (Mountain View, CA, USA). Subjects were asked to give basic demographic information and their patient code, which linked them to their anonymous data uploads.

We asked questions concerning the participant’s overall satisfaction and their adherence to regular patient inputs. Further inquiries concerned temporal effort of patient inputs, usability, design, and the willingness to use the Post OP Tracker app as a patient and as a health care professional (concerning only health care professionals).

For each of those questions, we gave the choice of four possible answers on a Likert scale, ranging from.

Additionally, we included questions concerning the technical operability of the app. We wanted to know if the notifications appeared regularly and at the desired time, and if there were any technical problems during the runtime of the study. Technical problems as well as a general feedback could be given using free text answers.

The complete questionnaire is given in Online Resource 1.

Participants reported an average satisfaction of 3.6/4. The app design was the lowest-rated entity on the questionnaire with an average score of 3.1/4. The willingness to use the Post OP Tracker app, both as a patient or as a health care professional, was rated highest with a score of 3.8/4 (Table 3).

Table 4 includes inquiries concerning technical aspects of the app. An important aspect of Post OP Tracker is the regularity of the notifications, requesting patient entry. Six users received the notifications regularly at the desired time (46.2%); one user did not (7.7%). The remaining six users who answered “other” (46.2%) were asked to further explain their answers. Some of those users reported a slight delay of a couple of minutes. Other users reported a 1-h delay after the change from daylight saving time to standard time. Indeed, this subject was overlooked by us when developing the app and concerned only users who started their runtime before the day of time change.

Ten users reported no technical issues (76.9%); two reported minor issues (15.4%). One user reported a missing notification at the end of the 12-week runtime, which would have requested him to make his last entry. Another user explained some uncertainties concerning the patient upload. The user who reported a major technical issue was the same person who answered the question concerning regularity of notifications with no. This person, who installed our app on a Huawei phone, was shown no notifications at all. It is a known issue that due to the strict settings concerning power savings, Huawei phones terminate background processes [28], in our case, the service showing the requesting patient entry notification.

We would like to illustrate how the Post OP Tracker app’s graphical results can be analyzed and interpreted.

Figure 5 shows an area graph of the simulated outcome of a participant. The y-axis shows the patient’s subjective well-being on a scale of 1 to 100; the x-axis the day of entry, ranging from 1 week before the operation to 12 weeks after the operation. This participant was tasked to simulate receiving a ventriculoperitoneal shunt operation for normal pressure hydrocephalus with a singular complication arising around 2 weeks after the surgery. The first decline in well-being marks the days preceding and immediately following the day of surgery, which may indicate anxiety besides progressive signs of hydrocephalus before and incisional pain from the wound after surgery. After this, the well-being increases as the patient recovers sharply but declines somewhat until the pre-defined complication arises. The complication was defined to be a generalized epileptic seizure that had to be treated with anticonvulsants (Clavien-Dindo grade II) in this scenario. The participant then simulated recovery from the complication in the following weeks.

Figure 6 shows the outcome of another participant undergoing the same surgery, however without any postoperative complications. Note how this participant chose to simulate a slower recovery than the participant in Fig. 5. However, in this case vignette, no complication occurred and, hence, no decline in the subjective well-being was recorded, as opposed to Fig. 5.

These two case vignettes contain individual and hence highly variable datasets. By gathering data from a high number of real patients after neurosurgery through the app, however, we could determine an “average normal postoperative course” for each type of surgical procedure without a postoperative complication.

In a second step, we could compare the outcome and postoperative course of patients who experienced any type of complication from those with an uneventful postoperative course. The difference between the “normal” and “abnormal” postoperative course could be calculated and its magnitude would be a measure of complication severity. This approach would take into consideration the impact of a complication on the patient’s well-being, other than the current classification systems.

For example, an imaginary 57-year-old female undergoing a shunt procedure for normal pressure hydrocephalus experiencing a postoperative urinary infection on day 5 (treated with antibiotics and total resolution 1 week later) will be classified as a CDG II type of complication. The same patient, now experiencing a small intraparenchymatous hemorrhage in the left basal ganglia, resulting in permanent hemiparesis and aphasia, will be classified as CDG I type of complication (as no treatment is required). The Post OP Tracker app will be able and differentiate between the severity of these complications better, as outlined in Fig. 7. Please note that the area between the average postoperative course and the individual postoperative course in patients with a postoperative complication can be used as an indicator of complication severity (marked in red color). With collected data with the help of this app, we could re-define and validate a new classification system for postoperative complications by ranking complications according to their impact on the patient’s subjective well-being.

A word of caution is required with regards to app-based data collection in patients that experience any type of complication resulting in cognitive decline or aphasia. Both complication sequelae would interfere with a patient’s ability to respond to the Post OP Tracker app’s pop-ups. Once we apply this app to real patients within the framework of an observational study, there are two options how to deal with this situation: First, if in doubt, the ability of a patient to continue with the study could be evaluated with cognitive screening tools, such as the Montreal Cognitive Assessment (MoCA) [16], a brief and effective tool in neurosurgery patients [20]. Patients with presumably permanent significant deficits will no longer be required to perform data entries and hence exit the study. However, as patients may recover over time, data entry could be temporarily halted and again continued after clinical improvement. Second, a patient’s next-of-kin might be approached to provide the missing data points during the time when the patient is unable to respond him-/herself. Both ways of dealing with this situation are sub-optimal from a scientific point of view but are pragmatic solutions that serve to estimate the impact of a given complication as accurately as possible.

We are aware that our current sample is limited (n = 13). The possibility of selection bias can also not be denied by the authors, since health care professionals and software developers are not necessarily reflective of the average neurosurgery patient population. Simulated “patients” were included in this feasibility study, who made their entries according to our pre-defined scenarios and their best knowledge. This, of course, is only an estimate of how patients, who experience real postoperative complications, will accept the app and classify outcome, especially during the onset and progression of neurological symptoms (e.g., postoperative headache, prolonged nausea or dizziness, neurological impairment, repetitive seizures …). However, we decided that this simplified approach is an ideal setting for testing a prototype app, without bothering patients and avoiding ethical issues. Technical operability has been established, as only one participant reported major technical issues, but this important feedback is helpful to improve the app, before applying it to patients. We chose the classic visual analog scale [21] as a way of representing the subjective functional status in order to minimize the temporal effort of patient inputs and thus ensuring patient adherence. However, more sophisticated patient-reported outcome measures (PROMs) of pain, disability, or quality of life could easily be implemented in the future. The exclusive availability of the Post OP Tracker app on Google Android phones is another limitation of our study. Further development should aim at an additional Apple iOS version, considering the iPhone’s market share of 43% in Switzerland [4]. A big strength is our active involvement in the development of our app, which allowed us to conserve financial resources and tailor Post OP Tracker precisely to our needs.