Total hip and knee arthroplasties are highly successful treatment modalities for advanced osteoarthritis, the most common joint disorder worldwide (Hiligsmann et al., 2013). A prosthetic joint infection (PJI), which develops in approximately 1 %–2 % of all arthroplasties, is a serious and devastating postoperative complication with a high impact on a patient's well-being (Kurtz et al., 2012; Zimmerli et al., 2004). Postoperative wound drainage is frequently reported as an important indicator for the presence of PJI (Wagenaar et al., 2019; Patel et al., 2007; Almeida et al., 2021). Wound drainage may be an early symptom of a present PJI but may also be a risk factor for the subsequent development of PJI (Patel et al., 2007; Weiss and Krackow, 1993). Discrimination between infectious and non-infectious postoperative wound drainage is of crucial importance. When the prosthetic joint is infected, surgical debridement and protracted antimicrobial treatment is required. For noninfectious serosanguinous drainage caused by intraoperative disruption of soft tissue and capillaries, only conservative wound management is indicated (Wagenaar et al., 2019).

In 2013, the first International Consensus Meeting (ICM) on PJI advised that surgical management of persistent wound drainage should be performed without delay if wound drainage persists for 5–7 d after index surgery (Parvizi et al., 2013). According to the recently published EBJIS (European Bone and Joint Infection Society) definition for PJI, a history of prolonged wound drainage (as a feature of wound healing problem) is a clinical sign included in the PJI likely category (Mcnally et al., 2021). However, these recommendations were not backed up by research data about duration of postoperative wound drainage as summarized in a recent systematic review (Wagenaar et al., 2019). Collecting wound drainage data is challenging because most patients are discharged from hospital soon after surgery. The use of smartphone applications (apps) for distant telemonitoring of postoperative patients has proven to be feasible and acceptable for both patients and surgeons (Sanger et al., 2014; Gray et al., 2010; Chua et al., 2017; Armstrong et al., 2017). In an earlier study, the use of a postoperative wound care app that was developed at the Leiden University Medical Centre showed a high perceived usefulness and ease of use as reported by patients (Scheper et al., 2019). To assess the amount and duration of postoperative wound drainage after joint arthroplasty in patients with and without PJI, we conducted a nationwide cohort study using this smartphone application in which we collected detailed information regarding the condition and natural history of the postoperative wound.

A multicentre, prospective observational study was conducted in 11 Dutch academic and non-academic hospitals between 1 November 2019 and 1 October 2021. All patients aged 18 years and older who received a knee or hip arthroplasty, who were able to provide informed consent, owned an Android or iOS smartphone and were able to read the Dutch language, were eligible for inclusion. Patients were screened during or after preoperative visits by a local nurse specialist or the coordinating study nurse. Informed consent was obtained via the wound care app. Instructions on how to use the app were provided to all patients by the local research coordinator. The nurses in the ward as well as the study coordinator were available for help with the use of the app during admission and throughout the study. All patients received routine postoperative medical care in the outpatient clinic as per local protocol in each participating hospital. The primary endpoint was the extent and duration of the postoperative wound drainage in patients with and without PJI. Secondary endpoints were the association between the presence of self-reported fever, redness, pain and PJI, and the validation of the designed algorithm for sending alert messages for suspected PJI. The PJI was defined according to the criteria from the European Bone and Joint Infection Society (EBJIS; Mcnally et al., 2021).

The use and function of the app has been described previously (Scheper et al., 2019). In short, for 30 d following joint arthroplasty, patients recorded their wound status daily on their mobile app. Redness, pain (by visual analogue score, VAS), wound drainage and presence of fever were recorded, and a picture of the wound could be taken. Based on the questionnaires, an inbuilt algorithm created a daily risk score (see Appendix A). If this score exceeded a predefined threshold, which was based on expert consensus of participating clinicians, an alert message was issued that allowed patients to contact their treating physician via a push button in the app. It was for the attending clinicians to decide whether patients needed a clinical review or not. If wound drainage during the first 14 d was not reported, patients were allowed to stop using the app. They were instructed to resume the use of the app if new drainage or other complications arose. After both 30 and 90 d, all patients were asked to report postoperative complications in the app. After a minimum follow-up period of 90 postoperative days, endpoint data were extracted from both the app and the electronic patient files to enable a comparison of the patient-reported and physician-reported outcome. If discordant, the outcome reported by the attending orthopaedic surgeon was regarded as the final outcome.

The study was conducted according to the principles of the Declaration of Helsinki. The study was approved by the ethics review committees and a waiver was obtained to use electronic instead of written informed consent. The use of the app for this study was approved by the Dutch Health Inspectorate (reference number VGR2O1 1434). The app was developed by the software company Innovattic. This company was not involved in the setup, data analysis and report of this study.

Of all patients eligible for inclusion during the study period, 1019 patients were included (total hip arthroplasty (THA) 46 %, total knee arthroplasty (TKA) 54 %). Baseline and outcome characteristics are summarized in Table 2. During the first 2 postoperative weeks, the app was used by more than 80 % of patients per day (Fig. 1). The app use declined during the third and fourth week from 80 % to 30 %, consistent with the recommendation that use of the app beyond 2 weeks was only needed if new drainage or other complications would occur.

The incidence of postoperative wound drainage in patients with and without PJI is reported in Table 3 and Fig. 2. During the first, second, third and fourth postoperative weeks, any form of wound drainage was present in 50 %, 12 %, 8 % and 3 % of patients without PJI and in 63 %, 88 %, 64 % and 25 % of patients with PJI. The high proportion of drainage in the first week was predominantly caused by minimal leakage (defined as <2×2 cm on gauze) occurring in 87 % (424/489) of patients without PJI. In this group, 51 patients (5 %) had moderate to heavy wound drainage in the first week, decreasing to 1 %, 1 % and 0.1 % in the next weeks. Moderate to heavy wound drainage of patients with PJI occurred in 25 %, 38 %, 46 % and 0 % of patients during 4 weeks. Reported redness (10 %), fever (5 %) and high pain scores (VAS >7, 11 %) were mainly reported during the first week and declined thereafter. Proportions of wound drainage in patients without PJI varied depending on the type of joint, BMI and the presence of diabetes (Table 4).

Sixteen (1.6 %) patients developed a PJI during the follow-up period. Fourteen patients experienced an early postoperative PJI after a median of 14 d (IQR 10–18 d). Two patients developed an early chronic PJI on postoperative day 71 and day 77 (Table 5). Three patients were reoperated for a suspected PJI that was subsequently not confirmed (e.g. haematoma). Six patients (0.6 %) received a short course of antibiotics for a presumed superficial wound infection but did not develop a PJI. The strongest risk factors for PJIs were any wound drainage in the second week (OR 50.83, 95 % CI 11.41–226.51) or moderate to heavy drainage in the second (OR 51.22, 95 % CI 15.84–165.65) or third week (OR 103.23, 95 % CI 26.08–408.57). New onset drainage in the second week after a week without drainage (OR 80.71, 95 % CI 9.12–714.52) and more than 5 cumulative wound drainage days during the first 3 postoperative weeks (OR 9.20, 95 % CI 3.37–25.14) were also strongly associated with the development of PJI (Table 6). Drainage for more than 5 d during the first 3 weeks predicted PJI with sensitivity of 63 % and specificity of 87 %, while drainage for more than 10 d predicted PJI with sensitivity of 27 % and specificity of 97 % (Appendix B). No wound drainage whatsoever was reported by 467 patients (46 %). Of them, only one patient developed a PJI resulting in a negative predictive value of no wound drainage as indicator for recovery without PJI of >98% (Table 6). The positive predictive value of any amount of wound drainage for PJI was low during the 4 postoperative weeks (2 %, 11 %, 8 % and 4 %, respectively) and increased for moderate–heavy wound drainage, especially in the third postoperative week (8 %, 35 %, 83 %, 0 %, respectively). Over the 4-week postoperative period, the average number of alerts per patient was not higher for patients with PJI compared to patients without PJI (OR 1.37 (0.39–4.87). Of the 18 420 d of app use, an alert was sent 2589 (14 %) times to 498 patients. In total, 141 (6.6 %) annotations could be obtained from the electronic patient files confirming that patients had contacted the hospital based on the sent alert. This led to a change of treatment in 61 (43 %) patients as summarized in Appendix C. Of the 16 patients who developed a PJI, an alert was sent in the preceding period to six patients which resulted in earlier outpatient evaluation or hospital admission in three patients.

Detailed knowledge of the extent and duration of wound drainage after arthroplasty is vital for orthopaedic surgeons who consider to reoperate patients with postoperative wound drainage for a suspected PJI. In this study, in which a mobile health application was used to monitor patients after arthroplasty, PJI was very unlikely in patients without any wound drainage. From the second week onward, wound drainage was strongly associated with the occurrence of PJI, but the sensitivity and positive predictive value of wound drainage as a single predictor for PJI was low. Due to the limited follow-up of 3 months, some patients with a late chronic PJI may have been missed. The insights from this study may help clinicians to evaluate postoperative patients who present with a leaking wound. Future research should focus on optimizing the algorithm, thereby improving the predictive value of the alert function.