Study characteristics

The literature consisted of 30,640 patients with OSA [19–30]. Ten studies used a comparative group of 317,085 patients without OSA [19–24,26–29]. There are nine hospital studies with 9088 patients with OSA and 122,936 patients without OSA [20,22–26,28,30]. There are six retrospective cohort hospital studies [20,22,24,26,28] and one prospective cohort study [23]. There are two prospective case-series studies [25,30]. Three studies were performed in ASCs with 21,455 patients with OSA and 193,545 patients without OSA [19,27,29]. All of these studies are cohort studies. One study was prospective [29]. There is one retrospective cohort study that used data from hospitals and ASCs with 97 patients with OSA and 604 patients without OSA [21]. Identification of OSA varied across studies: 4 used polysomnography data (PSG) [20,23,25,28], 10 relied on the preoperative record or self-report [19,21–24,26,27,29,30]. Notably, one study employed multiple methods of OSA identifications [23].

Unplanned hospital admissions

Unplanned hospital admissions for patients with OSA had a weighted average rate of 1.1%. (0.0–47.3%) [19–30]. Patients without OSA had a weighted average rate of 0.2% with less variation (<0.1–29.8%) [19–24,26–29]. ASC studies reported a 0.2% hospital admission rate [19,27,29] for patients with OSA and 0.1% unplanned admission rate for patients without OSA [19,27,29]. Three of these studies performed a statistical analysis and found no significant difference between OSA and non-OSA population [19,27,29]. Only one ASC study [19] used a selection protocol based on uncontrolled comorbidities and/or untreated sleep apnea. The other studies were assumed to rely on the surgeon/anaesthesiologist’s judgment [27,29]. Compared to ASC studies, outpatient hospital studies had higher rates of unplanned admissions for patients with OSA with a weighted average of 2.4% [20–26,28,30]. Patients without OSA had a unplanned admission rate of 0.1% in hospital studies [20,22–24,26,28]. Of these hospital studies, only three [20,23,25] used a selection protocol that took well controlled comorbidities and/or sleep apnea into consideration while the rest were up to anaesthesiologist/surgeon.

Six hospital studies statistically compared admission rates between the two groups and three found a statistically significant difference in admission rates [21,24,26]. Among these studies, Jackson et al. employed a broad selection protocol but did not use multivariate analysis to account for comorbidities as a potential cause of increased unplanned admissions [24]. Conversely, Masaracchia et al. is one of two large population studies within this review. It was the only study of its scale to incorporate a multivariate analysis considering age, sex, comorbidities, among other factors, and identified a statistically significant increase in unplanned admissions [26]. The only other study that had a population of that size was Melton et al. which incorporated a univariate analysis and found no significant difference in unplanned admissions. However, Melton et al was performed in the ASC setting.

24 hour postoperative complications

Five studies reported varied 24-hour postoperative complications rates [20,23,25,28,30]. All of these studies were performed in the hospital environment. One study estimated an all-complication rate of 16.0% for patients with OSA which was not statistically significant when compared to 13.1% in the non-OSA population [20]. Remaining studies reported cardiorespiratory complications for patients with OSA with a weighted average of 8.3% [23,25,28,30] with non-OSA controls having a complication rate of 12.2% [23,28].

The majority of 24 hour complications reported occurred during the PACU (within 4 hours of surgery) but 24 hour post-discharge reporting is rare among the studies. When severe complications within 24 hours were reported, patients without OSA had similar numbers of severe complications [20,28]. Moreover, reporting of cardiorespiratory complications varied across studies. Kleipool et al. only reported unstable vital signs that led to admissions, assumed as cardiorespiratory complications [25]. Hudson et al. reported all postoperative cardiorespiratory 24 hour complications [23].

30 day postoperative complications

Thirty day complications primarily consisted of post-discharge readmissions and revisits. Only three studies reported an overall 30-day complication rate [25,26,30]. Kleipool reports a rate of 6.1% complication rate for its population [25]. Masaracchia et al. report a significantly higher complication rate of 4.10% in the OSA population compared to 1.22% in the non-OSA population [26]. Masaracchia et al. also reported a higher rate of pulmonary compromise, myocardial infarction, pneumonia, and acute renal failure in the OSA population. Readmission rate was reported in 5 studies and had a weighted average rate of 0.3% (0.0–8.2%) [20,23,25,26]. Only one study reported ER revisits which gave a rate of 14.3% [25]. Among the studies that reported on mortality rates, only Masaracchia et al., a large population study, documented a single death in the OSA population and found no statistically significant difference in mortality rates between the OSA and non-OSA populations [20,22,23,25,29,30].

Risk of bias

Tables 9 and 10 present the risk of bias for included studies. (S3 Data) 9 studies were analyzed using the NOS for risk of bias analysis [19–24,26–28]. Nine studies reported a 6 or 7 on the NOS with an average of 6.6, a moderate quality score. Almost all studies reported a 0/2 on comparability [19,20,22–24,27]. While many of these studies reported age, BMI, and/or AHI, only three studies considered confounding factors in its design/analysis [21,26,28]. Sabers et al. used a case-control design that considers gender, age, and BMI but not comorbidities and severity of OSA. Masaracchia et al. used a multivariate analysis that considers gender, age, BMI and comorbidities but not severity of OSA. 3 studies were analyzed using Murad’s tools [25,29,30]. Two studies were strong in all areas while Stierer et al. had weak ascertainment and selection domains.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 10. Risk of bias tool using Newcastle-Ottawa scale.

https://doi.org/10.1371/journal.pone.0326704.t010

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 9. Risk of bias tool for case series using murad tool.

https://doi.org/10.1371/journal.pone.0326704.t009

Unplanned admissions

Similar to Ceban et al., this review found that patients with OSA undergoing ambulatory surgery are at a higher risk of unplanned admission. While smaller population studies do not find this trend, large scale studies within this review have found significant increases of unplanned admissions [21,24,26]. Jackson et al., which employs a broad selection protocol, has shown that patients with OSA are more likely to have unplanned admissions. However, this is likely secondary to the high prevalence of comorbidities (diabetes, hypertension, obesity) in patients with OSA [31–33]. More narrow selection protocols within large population hospital studies continue to report significant increases in unplanned admissions even when accounting comorbidities with a multivariate analysis [21,26]. There are multiple physiological factors for this. The low pharyngeal tone of OSA makes these patients vulnerable to airway collapse after the use of anesthetics, opioids, and sedation [34]. Furthermore, OSA has been associated with hemodynamic instability post-operatively [7].

However, ASC studies included within our review did not find similar results. Melton et al. includes over 21 thousand patients with OSA and did not report a statistically significant higher risk of unplanned admission [27]. Furthermore, Bryson et al, a hospital study, had a similar number of patients with OSA in comparison to Christian et al. and still found no significant difference in unplanned admission [20,21]. Several factors may explain this discrepancy within our data. First, Christian et al. and Masaracchia et al. both were orthopedic surgery studies. Ceban et al found that orthopedic surgeries are associated with a higher rate of unplanned admission in comparison to other ambulatory surgeries. Melton et al and Bryson et al. included a broader range of surgeries. So, patients with OSA were less likely to undergo invasive and painful surgeries within their studies. Second, ASC studies like Melton et al. will likely have healthier OSA patients in comparison to hospital settings.. Silber et al. found similar results within their study comparing elderly patient undergoing surgery at ASCs and hospital settings [35].

Postoperative complications

All studies that reported 24 hours complications were hospital studies. However, for every study that reported severe complications in patients, a similar rate of severe complications in 24 hours. was found in the patients without OSA. Masaracchia et al reported multiple severe complications are more likely to occur in the ambulatory surgery patient population 30 days after surgery. However, it should be noted that Masaracchia et al. get its data from hospitals and assume that all shoulder arthroplasty will be performed ambulatory. This will mean this data includes patients that were not intended to undergo outpatient surgery and more likely to have severe complications.

With appropriate patient selection and similar rates of severe complications within 24 hours, patients with OSA can safely undergo surgery in ASC environments. Patient selection is crucial to ensuring the safety of patients with OSA in ASCs. As demonstrated by Masaracchia et al., OSA remains an independent risk factor for admissions; therefore, surgeons and anesthesiologists should maintain more selective criteria when assessing comorbidities and patient characteristics. Patients with OSA selected for outpatient surgery at an ASC should have well-managed comorbidities,treated sleep apnea, and with no severe cardiovascular, respiratory, or metabolic diseases. Furthermore, it is advisable to implement perioperative interventions, including the use of CPAP machines post-operative, opioid-sparing techniques, and thorough preoperative screening for undiagnosed sleep apnea [36,37]. Additionally, it is crucial to consider specific patient characteristics associated with OSA. The literature indicates that morbid obesity and age over 65 are significant risk factors for unplanned admissions following ambulatory surgery [38,39].

Concerns of undiagnosed sleep apnea exist within this data. Literature has estimated that about 90% of sleep apnea are undiagnosed and undiagnosed sleep apnea could be a cause for why large sample sizes are needed to show differences in unplanned admissions. Bryson et al. and Hudson et al. control for this factor by using prior PSGs and using preoperative respiratory polygraphy to control for undiagnosed sleep apnea, respectively [20,23]. However, both studies were small sample size studies and found OSA did not increase unplanned admissions.

Our review has several limitations. Firstly, the one most common complication of OSA patients undergoing surgery was oxygen desaturation in PACU. However, these patients have been desaturating nightly for years pre-operatively, using this as a complication measure is likely inflating the complication rate. Secondly, there was no limitations within our inclusion criteria on selection/discharge protocol. As a result of this, we included many studies such as Jackson et al. that were not relevant to the ASC environment. Furthermore, our review lacked a meta-analysis for a more quantitative assessment of the evidence. Without data aggregation across studies, our data is less robust and should be interpreted with due caution. Further large population research is required within ASCs so that robust meta-analysis can be performed on this topic. Another significant limitation of our review was the predominance of retrospective studies which are prone to biases and errors such as selection bias. Moreover, complication reporting in many studies had unclear or incomplete data, leading to the use of surrogate measures like indications for unplanned admissions, introducing uncertainty. Additionally, it is unclear whether 30-day readmissions were a direct consequence of the surgery itself or unrelated causes and underreporting may exist as most studies only retrieve data from one hospital, overlooking patients who may have been readmitted elsewhere. The absence of detailed, and consistent complications reporting limits the reliability of our findings. The heterogeneity across study methodologies, patient populations, and surgical procedures complicates the task of drawing firm conclusions and highlights the need for standardized approach in future research. Lastly, our review is limited by the studies that are currently available; unpublished studies or unindexed studies could alter our conclusions.