Perioperative and anesthesia-related cardiac arrests in geriatric patients: a systematic review using meta-regression analysis

The worldwide population is aging, and the number of surgeries performed in geriatric patients is increasing. This systematic review evaluated anesthetic procedures to assess global data on perioperative and anesthesia-related cardiac arrest (CA) rates in geriatric surgical patients. Available data on perioperative and anesthesia-related CA rates over time and by the country’s Human Development Index (HDI) were evaluated by meta-regression, and a pooled analysis of proportions was used to compare perioperative and anesthesia-related CA rates by HDI and time period. The meta-regression showed that perioperative CA rates did not change significantly over time or by HDI, whereas anesthesia-related CA rates decreased over time (P = 0.04) and in high-HDI (P = 0.015). Perioperative and anesthesia-related CA rates per 10,000 anesthetic procedures declined in high-HDI, from 38.6 before the 1990s to 7.7 from 1990–2017 (P < 0.001) and from 9.2 before the 1990s to 1.3 from 1990–2017 (P < 0.001), respectively. The perioperative CA rate from 1990–2017 was higher in low-HDI than in high-HDI countries (P < 0.001). Hence, a reduction in anesthesia-related CA rates over time was observed. Both perioperative and anesthesia-related CA rates only decreased with a high-HDI between time periods, and perioperative CA rates during 1990–2017 were 4-fold higher with low- compared to high-HDI in geriatric patients.

regarding worldwide rates of perioperative and anesthesia-related CA events in geriatric patients is available in the literature. We hypothesized that perioperative and anesthesia-related CA rates have decreased over time in high-and low-HDI countries by performing a weighted meta-regression and weighted CA rates for geriatric patients published in cohort and cross-sectional studies.
The current study aimed to perform the first meta-regression analysis of the worldwide data on perioperative and anesthesia-related CA rates in geriatric patients over time and according to national HDI values. Additionally, we examined pooled weighted CA rates of perioperative and anesthesia-related CA events in low-and high-HDI countries during the pre-1990s and the 1990-2017 period.

Results
Selection of studies. Our literature search yielded 20,074 citations in addition to 71 potential studies that were identified in the references of related articles. We excluded 2,648 duplicate studies, and after reviewing the titles and abstracts, we retrieved 290 potentially relevant full-text articles for more detailed evaluation. Of these articles, 16 studies met the inclusion criteria ( Fig. 1).

Study characteristics.
Sixteen studies from nine countries fulfilled the inclusion criteria. These studies involved 1,758,153 anesthetic procedures performed during surgery in geriatric patients. The characteristics and designs of these studies are listed in Table 1.

Meta-regression analysis.
Of the evaluated studies (n = 16), 13 were related to perioperative CA (Figs 2 and 3), and eight were related to anesthesia-related CA (Figs 4 and 5).

Time (study year).
The data from all of countries assessed in the weighted meta-regression showed no statistically significant relationship between perioperative CA and time (slope: 0.0079; 95% CI: −0.0483 to 0.0642; P = 0.76; Fig. 2). However, there was a significant decrease in the anesthesia-related CA rate over time, regardless of HDI status (slope: −0.0699; 95% CI: −0.1394 to −0.0003; P = 0.04; Fig. 4A). There was also a significant decrease in this rate over time in high-HDI countries (slope: −0.1049; 95% CI: −0.1762 to −0.0336; P = 0.015; Fig. 4B). Correlation between anesthesia-related CA rates and time was not calculated for low-HDI countries because these rates were described in only two articles, which made the meta-regression analysis impossible.
Pooled weighted analysis of CA rate. When we compared studies to determine whether a country's HDI status had an effect on CA events during the two evaluated time periods (pre-1990s versus 1990-2017), we found that in high-HDI countries, perioperative CA rates decreased from 38.6 (95% CI: 37.8-39.6) per 10,000 procedures before the 1990s to 7.7 (95% CI: 7.6-7.8) per 10,000 procedures in 1990-2017 (P < 0.001); anesthesia-related CA rates decreased from 9.2 (95% CI: 8.9-9.5) per 10,000 procedures before the 1990s to 1.3 (95% CI: 0.2-6.5) per 10,000 procedures in the 1990-2017 (P < 0.001; Table 2). None of the studies evaluated perioperative or anesthesia-related CA rates in low-income countries before the 1990s, making sub-analysis of data for these periods impossible.
For all event rates, the proportion of heterogeneity (I 2 ) varied from 71.8% to 98.9% (Table 2).

Discussion
In this systematic review of surgeries in geriatric patients, we used meta-regression analysis to show that perioperative CA rates did not change over time or with regard to HDI status; in contrast, anesthesia-related CA rates did significantly decrease over time but not according to HDI status. In addition, by using the pooled weighted CA rate analyses, we found that both perioperative and anesthesia-related CA rates were lower in high-HDI countries in the later time period. However, it was not possible to analyze changes in these event rates in the low-HDI countries because there was a lack of published studies covering the period before the 1990s. During the 1990-2017 period, perioperative CA rates were 4-fold lower in high-HDI than in low-HDI countries. Based on these findings, the following question arose: Why did perioperative CA rates not decline over time or by HDI status in geriatric patients? A previous meta-regression analysis of data obtained from a systematic review of anesthesia-related and perioperative mortalities in high-and low-income countries showed that the baseline American Society of Anesthesiologists (ASA) physical status of patients has increased significantly in recent  14 . In the same study, meta-regression by ASA status was performed to evaluate CA events and deaths, and a significantly increasing relationship was found between a higher ASA status at baseline and CA events or deaths 14 . Additionally, studies evaluating low-and high-HDI countries have demonstrated that geriatric patients who present a poorer ASA physical status (III-V) are more likely than ASA I-II patients to suffer perioperative CA or death [16][17][18][19] . Elderly individuals are more predisposed than young people to cardiovascular disease. For example, a ruptured aneurysm or complications associated with cardiac surgery or intraoperative myocardial infarction are important causes of intraoperative CA and death in geriatric patients 16 . Therefore, the baseline risk for and complexity of managing geriatric patients undergoing surgery are both increased. Geriatric patients, many of whom are frail, can experience considerable comorbidities and cognitive impairment and have reduced physical activity, all of which are associated with perioperative complications 20 . These factors are thought to contribute to the significant increasing correlation between perioperative CA rates and time. However, we found no significant change in this relationship, possibly indicating that increased perioperative safety measures are implemented with regard to geriatric patients.
The significant reduction we observed in anesthesia-related CA rates over time may be associated with considerable improvements in anesthesia safety since the early 1990s, especially in high-HDI countries. We can also attribute this finding to several safety improvements, such as advancement in drugs, training programs, guidelines and checklists, investments in monitoring techniques, especially the capnograph and pulse oximeter, and systematic approaches to reduce errors [21][22][23] . According to Eichhorn, these initiatives, which are aimed at achieving "safety monitoring", have resulted in a continuous increase in intraoperative vigilance and improved patient safety during anesthesia 24 . In our study, the relationship between anesthesia-related CA rates and time could not be evaluated in low-income countries because only two relevant articles were identified. This made performing meta-regression analysis using these data impossible.
The results of our pooled weighted CA rate analyses, which were divided into HDI status and time period, demonstrated the existence of a considerable gap between the healthcare systems of low-and high-income  countries. For example, in high-HDI countries, there were significant 5-and 7-fold reductions in perioperative and anesthesia-related CA rates, respectively, between the two time periods. Additionally, our study showed that the perioperative CA rate was 4-fold lower in high-HDI than in low-HDI countries from 1990 to 2017. Although the anesthesia-related CA rate was also 3.5-fold higher in low-HDI countries than in high-HDI countries during this period, this difference was not significant. Regardless, this result may have been influenced by the relatively small number of both studies and patients that were identified for low-HDI countries.
Several studies have highlighted the major role that pre-anesthetic management of a patient's condition plays in minimizing complications and adverse outcomes 18, 25-28 . One study performed in a low-income country showed that several geriatric patients have poor health when undergoing surgery 16 . These findings demonstrate that there is a need for improving the quality and quantity of resources that can be used as well as access to healthcare, both of which are inadequate, in developing countries. Additionally, it is necessary to adopt  perioperative medical practices that have demonstrable effectiveness, to organize multidisciplinary discussions of adverse effects, to implement evidence-based safety protocols, to provide better methods for selecting patients for surgery, and to initiate advances in techniques, protocols, and pre-and postoperative critical care to improve perioperative patient care 29,30 .
There are some data limitations in the current study. One of the main limitations is that several papers were excluded from the review due to the lack of information about both the numbers of geriatric patients and the patients who developed CA in this specific age group. Additionally, many of the included papers presented different age group subdivisions, including both adult and geriatric patients who developed CA (50-69 8 ; 15-69 31 ; 55-74 years 3,4 ); this only enabled the inclusion and analysis of patients ≥70 or ≥75 years, respectively, in the aforementioned studies. There were also differences across the included studies regarding surgical populations (e.g., excluded ASA V patients), time frames (e.g., intraoperative, 24 hours or 48 hours postoperatively, or seven postoperative days), and procedures (e.g., exclusion of organ transplantation, trauma, or cardiac surgery). Accordingly, these differences contributed to the high heterogeneity observed in our study. A random-effects model was applied to minimize this heterogeneity when assessing trends between the two time periods and the two HDI status categories. In addition to including ≥3,000 geriatric patients, we also calculated the weighted rate of CA events across all studies to minimize possible bias. The last limitation is that there were no pre-1990s studies performed in low-income countries that either fulfilled the inclusion criteria or were not published in an indexed journal, and the absence of such studies may have contributed to publication bias.
We evaluated a public health problem that is associated with the geriatric population. This age group experiences a considerable number of diseases and functional dependencies that contribute to higher health system costs and a substantial impact on economic, social, and family dynamics 32 . It is therefore necessary to expand and intensify our research into aging and evidence-based medicine to provide better geriatric care in high-HDI and, in particular, low-HDI countries 33 . Indeed, there is a large disparity between the healthcare systems of these countries, and our study showed that improvements have been made in geriatric patient safety care during anesthetic procedures in high-but not in low-income countries. As explained by Koga et al., future efforts and collaborations should involve improvements in perioperative safety to reduce the gap between high-and low-HDI countries in their respective health care systems 15 . Further reviews of perioperative and especially anesthesia-related CA must be periodically performed to obtain continued worldwide CA rates in geriatric patients in low-and high-HDI countries. We also showed that there was a decrease in anesthesia-related CA rates over time while there was an increase in the size of the geriatric population and the associated number of comorbidities.
In conclusion, our meta-regression analysis shows that perioperative CA rates did not change over time or based on a country's HDI; conversely, anesthesia-related CA rates decreased over time but not according to HDI. By using pooled weighted CA rate analyses, a notable reduction in perioperative and anesthesia-related CA rates between the two time periods was observed in high-HDI countries. In addition, perioperative CA rates were 4-fold lower in high-than in low-income countries during the 1990-2017 period.

Methods
This manuscript was prepared in accordance with Meta-analyses Of Observational Studies in Epidemiology statements (MOOSE) 34 . According to the local institutional review board (IRB), there was no need for Ethical approval due to the type of study.

Search Strategy and Inclusion and Exclusion Criteria.
The following databases were searched until April 2 nd , 2017: EMBASE, MEDLINE, LILACS, and SCIELO. Additionally, we manually reviewed all potential studies and included relevant articles. No restrictions were applied for year of publication or language. The literature search was performed using MeSH terms and text words including synonyms about perioperative and/or anesthesia-related fatal and non-fatal CA rates in geriatric patients. To achieve higher sensitivity and to identify relevant articles, the search strategy was adjusted for each database (see Supplementary S1.1 and S1.2).
The following inclusion criteria were applied for the studies: (i) included patients aged 60 years and older and reported perioperative and/or anesthesia-related CA rates; (ii) involved all age groups, including patients aged 60 years old and older; (iii) a cohort or cross-sectional study; (iv) specified perioperative and/or anesthesia-related CA rates that occurred seven days after the surgery; and (v) provided enough information to perform the analysis.
The exclusion criteria were as follows: (i) reported only one surgical procedure (e.g., cardiac surgery) or a specific anesthetic technique (e.g., regional anesthesia) or patient subtype (e.g., patients with ASA physical status I and II only); (ii) did not specify the time period; or (iii) evaluated fewer than 3,000 geriatric patients. At least 3,000 patients in each study were included to allow us to estimate the incidence of rare adverse events (≤1 per 1,000 anesthetics) 35 . The primary outcome was perioperative CA (an event resulting from any cause; e.g., patient disease/condition, surgery, or anesthesia). The secondary outcome was anesthesia-related CA (either partly or completely attributable to anesthesia) by the investigators of the included studies. Because there is no consensus regarding the definition of the postoperative period (in the worldwide literature, the definitions of postoperative period include the first 24 hours, 48 hours, and even 1 month after surgery), the period of perioperative CA was defined in this review as the date of surgery until the seventh postoperative day.

Data Extraction and Outcome
High-and low-income countries were defined as having an HDI ≥ 0.8 and <0.8, respectively. A country's HDI status can change over time, and several studies have reported data covering a time period of many years; thus, HDI was defined as the mean of HDI values between the first and last year in which the patients were recruited for each study. When HDI was not available for the specific time period covered by the study, we used the HDI for the closest available date.
Due to the fact that many safety-improvement measures emerged in the early 1990s in high-HDI countries and later in some low-HDI countries and based on two systematic reviews of perioperative CA 14,15 , the time frame evaluated in the studies included in this analysis was split into two periods (pre-1990s versus 1990-2017). These measures included improvements in the care of patients, which involved equipment sterilization, new anesthesia medications and workstations with ventilators, anesthesia-practice based on safety protocols, and an increase in the number of adult intensive care beds 14, 15, 21-23 . Statistical Analysis. Our meta-regression analysis used a fixed-effect model with restricted estimated maximum likelihood (REML) and an observed log-odds ratio to predict whether CA rates in geriatric patients statistically changed over time or according to HDI (time and HDI were evaluated as continuous variables). Stata-13 (StataCorp LP, College Station, TX) was used to perform the meta-regression.
Additionally, a random-effects model was applied to calculate weighted event rates across all of the included studies with a pooled analysis (StatsDirect Ltd, Altrincham, Cheshire, UK) 36 . The times and HDI values were divided (i.e., pre-1990s versus 1990-2017 and low-income versus high-income, respectively) to evaluate perioperative and anesthesia-related CA rates. The event rate was defined as the number of CA events per 10,000 anesthetic procedures, and the data are reported as 95% confidence intervals (CIs). Proportion tests were performed using SAS for Windows ® , v.9.4 (SAS Institute, Cary, NC). Chi-square tests were used to compare differences in the proportions of events according to time period and HDI status. Because the data were recorded as time intervals (e.g., from 06/06/1989 to 06/06/1993), the median year of the study's patient recruitment interval (i.e., median: 1991) 15 was considered. A P value < 0.05 was considered significant.
An alternative approach that is employed to quantify the effect of heterogeneity is the I 2 statistic (StatsDirect Ltd, Altrincham, Cheshire, UK) 37,38 ; values higher than 50% suggest heterogeneity among the studies.