Background

As the number of older adults increases globally, there will be a greater need for elective surgeries in this patient population; however, elective surgeries can be associated with significant harm to patients [1,2,3,4,5]. Special preoperative consideration must be given to the greater prevalence of geriatric syndromes faced by older adults, such as frailty and functional impairment, that potentially increase their risk of adverse postoperative outcomes [6, 7]. Indeed, older adults are a heterogeneous group of patients whose risk of adverse postoperative outcomes is not adequately described by chronological age, comorbidities, or the type of surgical procedure alone [8]. Although older adults are often seen in the preoperative medicine clinic for cardiovascular and respiratory risk stratification and optimization in anticipation of an elective surgery, little consideration is given to risk stratification for other adverse outcomes that occur in older adults, despite the availability of information to aid in this assessment [9].

Understanding the risk factors for postoperative complications may help clinicians, patients, and caregivers to target non-pharmacological and pharmacological interventions aimed at lessening the burden of these adverse postoperative outcomes. This systematic review synthesizes studies that identify preoperative prognostic factors of older adults undergoing elective surgery which may predispose them to adverse postoperative outcomes. This information can be used by clinicians and patients to enhance decision-making and management in the preoperative setting and by researchers to study possible interventions aimed at improving postoperative outcomes for older adults.

Methods

This study was reported in accordance with both the PRISMA statement for reporting systematic reviews and meta-analyses and the MOOSE statement for reporting meta-analysis of observation studies in epidemiology (Additional file 1) [10, 11]. This systematic review and meta-analysis has a companion publication that focuses on prognostic factors associated with postoperative delirium among older adults undergoing elective surgery.

Eligibility criteria

Prospective studies (e.g., randomized controlled trials (RCTs), quasi-RCTs, non-RCTs, controlled-before-and-after studies, prospective cohort studies) were eligible if they included older adults undergoing elective surgery (≥60 years old and mean age of patients enrolled in the study ≥ 65 years old) and reported prognostic factors associated with the postoperative complications of mortality, functional decline, prolonged length of hospitalization, discharge to a location other than home, and a composite outcome of medical or surgical complications. All definitions of a given prognostic factor were included. Studies that included patients ≥ 60 years old were selected to align with definitions from the United Nations and the World Health Organization [12, 13]. Geriatric medicine consultation services typically target these age ranges [14, 15]. Studies using any method for diagnosing postoperative complications were eligible. Postoperative mortality was defined as death within 30 days following surgery. If a study reported both elective and emergent surgical procedures, it was included in our systematic review only if there was a separate subgroup reported for patients undergoing elective surgery. To make the review feasible, studies reporting only clinical, laboratory, or imaging investigations that are not conducted as part of routine clinical practice (i.e., measuring serum interleukin levels) were excluded, as were studies disseminated in languages other than English.

Information sources and search strategy

An experienced librarian searched MEDLINE (OVID interface, 1948 to April Week 3, 2016), EMBASE (OVID interface, 1980 to April Week 3, 2016), CINAHL (EBSCO interface, 1994 to April 21, 2016), Cochrane Central Register of Controlled Trials (Issue 4, April 2016), and AgeLine (EBSCO interface, 1968 to April 21, 2016) for potentially relevant studies. The full search strategy for MEDLINE (Additional file 2: Appendix 1) was modified as necessary for the other databases (full searches available upon request). Scanning the reference lists of included studies and searching the authors’ personal files supplemented the electronic search. Authors of conference proceedings were contacted to obtain unpublished work.

Study selection

Two levels of screening were completed independently by two reviewers using Synthesi.SR (proprietary online software developed by the Knowledge Translation Program, Toronto, Canada); these were level 1 screening of titles and abstracts, and level 2, full-text screening of articles. A calibration exercise was conducted prior to level 1 screening whereby each reviewer independently screened 10% of a random sample of citations to ensure adequate inter-rater agreement. Study authors were contacted for further information if it was unclear whether the study met inclusion criteria. Disagreements concerning article inclusion were resolved through discussion; otherwise, a third reviewer was available to make a final decision.

Data abstraction

Data were abstracted independently by two reviewers from studies retained from level 2 screening. Study characteristics (e.g., study design, country of conduct), patient characteristics (e.g., mean age, sex, comorbidities), and prognostic factors associated with the outcomes of interest were abstracted from included studies. Definitions operationalized by study authors for individual prognostic factors were also abstracted, where appropriate. Conflicts regarding the abstracted data were resolved through discussion. Authors were contacted for further information when the data were not clearly reported. The publication with the longest duration of follow-up was considered the major publication when multiple studies reported data from the same source. The other publications were retained as supplementary material only.

Methodological quality assessment

Two reviewers independently appraised the risk of bias using the Cochrane Risk-of-Bias Tool for RCTs and the Newcastle–Ottawa Scale for cohort studies [16, 17]. We planned to assess other study designs with the Cochrane Effective Practice and Organization Care (EPOC) Risk-of-Bias Tool [18]. If two or more outcomes were reported in a single study, the quality assessment was preferentially conducted on the outcome of postoperative complications or destination at discharge from hospital.

Statistical methods

We calculated odds ratios (OR) to quantify the relative risk of postoperative complications associated with each prognostic factor. Whenever only continuous effect measures, such as mean differences (e.g., age, body mass index) were reported, these effect sizes were transformed to OR estimates, if needed, to derive an overall effect estimate that combined both dichotomous and continuous study-level effect estimates [19]. For studies that reported multiple options with which to derive the study-level effect estimate (e.g., 2 × 2 tables, adjusted and unadjusted ORs, mean differences), the order of preference for selecting the source data is described in Additional file 2: Appendix 2.

Random effects models were used to derive overall effect estimates with 95% confidence intervals (CIs) when two or more studies reported extractable effect estimates that could be combined for the purpose of meta-analysis. The number needed to follow (NNF) was calculated as 1/pooled incidence of each postoperative complication. Similar to the concept of the number needed to treat or number needed to harm, the NNF represents the number of patients who need to be followed in a prognostic study in order to see one outcome [20]. Information regarding data imputation methods to approximate standard deviation values is found in Additional file 2: Appendix 3. Between-study statistical heterogeneity was quantitatively assessed with the I2 statistic and thresholds for the interpretation of the I2 statistic were consistent with those reported in the Cochrane Handbook for Systematic Reviews of Interventions [21].

Subgroup analyses were conducted by surgery type to explore between-study heterogeneity. Mixed-effects meta-regression models were also used to evaluate the effect of study-level effect modifiers (age, publication year, and type of surgery) on the pooled incidence of postoperative complications. Sensitivity analyses were conducted based on the type of study-level effect estimates used to calculate the overall effect estimates, - including study-level effect estimates that were adjusted for potentially important confounders only. A prognostic factor was considered significantly associated with the primary or secondary outcomes at a two-tailed p-value < 0.05. We planned to test for publication bias; however, this was not possible because there were no prognostic factors that were reported in at least 10 studies. All statistical analyses were conducted in R, version 3.2.4, using the metafor and meta packages [22, 23].

Results

Of the 5692 titles and abstracts that were screened for inclusion, 44 studies, including 12,281 patients, met our inclusion criteria (Fig. 1). From the 44 included studies, prognostic factors associated with postoperative complications (n = 28), postoperative mortality (n = 10), length of hospitalization (n = 22), functional decline (n = 6), and destination at discharge from hospital (n = 13) were retrieved. Two RCTs were included, which were at moderate to high risk of bias (Additional file 2: Appendix 4). Overall, the included cohort studies were of moderate to high methodological quality (Additional file 2: Appendix 5). The most common biases were the adequacy of follow-up of cohorts and the comparability of the cohorts on the basis of design.

Fig. 1
figure 1

Study flow

Full size image

Postoperative complications

Twenty-eight studies (6708 patients) investigated the association between preoperative prognostic factors and postoperative complications (Additional file 1: Appendix 6). Of these, 23 were included in the meta-analyses of prognostic factors [1, 6, 7, 24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43]. The five studies not included in meta-analyses did not contain extractable data, report prognostic factors that were included in two or more studies, or present data in a format that could be pooled with other study-level effect estimates. Postoperative complications were most often reported as a composite of postoperative medical or surgical complications (e.g., pneumonia, wound infection, venous thromboembolism), the details of which are found in Additional file 2: Appendix 6. The pooled incidence of postoperative complications across all surgical types was 25.16% (95% CI 18.26–33.61%, 21 studies, I2 = 96%, NNF = 4) [1, 7, 25,26,27,28, 30,31,32,33, 35,36,37,38, 40,41,42,43,44,45,46]. In exploring the influence of type of surgery on incident complications, the number of complications remained high: cardiac surgery (9.46%, 95% CI 2.71–28.18%, 3 studies, I2 = 96.40%, NNF = 11), abdominal surgery (24.73%, 95% CI 8.63–53.33%, 3 studies, I2 = 96.1%, NNF = 5), and thoracic surgery (33.97%, 95% CI 12.66–64.62%, 4 studies, I2 = 95.5%, NNF = 3). The effects of the mean age of study patients, publication year, and type of surgery on the pooled incidence of postoperative complications were explored with meta-regression, but did not explain any of the variance in the models.

The prognostic factors most strongly associated with the development of postoperative complications were poor performance status as defined by the Eastern Cooperative Oncology Group (ECOG) score or the Karnofsky Index (OR 2.58, 95% CI 1.56–4.25, 5 studies, I2 = 0%), smoking status (OR 2.43, 95% CI 1.32–4.46, 3 studies, I2 = 0%), impairment in instrumental activities of daily living (IADLs) (OR 2.27, 95% CI 1.65–3.14, 6 studies, I2 = 0%), frailty (OR 2.16, 95% CI 1.29–3.62, 8 studies, I2 = 54.69%), and cognitive impairment (OR 2.01, 95% CI 1.44–2.81, 8 studies, I2 = 0%) (Table 1, Additional file 2: Appendix 7). Frailty was most frequently defined using the definition of Fried et al. [47]; however, other definitions included the Edmonton Frailty Scale, gait speed, or a tool created by individual study authors [48]. In a subgroup of frail patients undergoing abdominal surgery, there was no longer an association between frailty and postoperative complications (OR 1.73, 95% CI 0.81–3.66, 3 studies, I2 = 53.36%) [29, 35, 40]. These findings remained consistent when sensitivity analyses were conducted whereby only those studies reporting study-level effect estimates adjusted for important confounders were included.

Table 1 Prognostic factors for postoperative complications among older adults undergoing elective surgery
Full size table

Other prognostic factors that were reported in single studies as significantly associated with postoperative complications were the cumulative number of impairments in the comprehensive geriatric assessment (OR 1.84, 95% CI 1.27–2.65), not being able to shop independently (P = 0.011), answering ‘yes’ to the question ‘Have you dropped many of your activities and interests?’ on the Geriatric Depression Scale (GDS) [49] (P = 0.04), the presence of one or more Goldman indicators [50] (P < 0.005), and the inability to bicycle 2 minutes to a heart rate greater than 99 beats/min (P < 0.05) [44, 45, 51]. The presence of anxiety (OR 5.1, 95% CI 1.27–20.2), Society of Thoracic Surgeons score [52] (OR 1.06, 95% CI 1.01–1.10), and female sex (OR 3.49, 95% CI 1.52–7.99) were associated with mortality or major morbidity in patients undergoing cardiac surgery [53].

Energy intake > 21.3 kcal/kg of actual body weight (OR 2.40, 95% CI 0.59–9.80), energy intake > 22.2 kcal/kg of ideal body weight (OR 5.00, 95% CI 0.95–26.17), or any of the items on the Nutrition Screening Initiative Nutritional Health Checklist [54] were not associated with postoperative complications [45, 46]. Besides answering ‘yes’ to the question ‘Have you dropped many of your activities and interests?’ on the GDS, none of the other questions were associated with postoperative complications. Similarly, no activities of daily living (ADLs) or IADLs, besides shopping, were individually associated with postoperative complications [45].

Postoperative mortality

The association between preoperative prognostic factors and postoperative mortality was investigated in 11 studies (3399 patients) (Table 2) [1,2,3, 25, 30, 32, 51, 53, 55,56,57]. The pooled incidence of mortality was 4.58% (95% CI 3.67–5.71%, 11 studies, I2 = 46.30%, NNF = 21) [1,2,3, 25, 30, 32, 51, 53, 55,56,57]. Among patients undergoing cardiac surgery, the pooled incidence of mortality was 5.21% (4.00–6.75%, 6 studies, I2 = 60.8%, NNF = 20). Only the effects of publication year could be explored in a meta-regression because there were not enough studies to explore the effects of type of surgery or mean age of patients on the pooled incidence of mortality. Publication year did not explain any of the variance in the meta-regression model. Few prognostic factors were reported in more than one study. No significant association was identified between male sex (OR 1.46, 95% CI 0.67–3.19, 4 studies, I2 = 53.92%), diabetes mellitus (HR 1.74, 95% CI 0.54–5.61, 2 studies, I2 = 45.26%), or history of heart failure (HR 1.86, 95% CI 0.44–7.88, 2 studies, I2 = 68.34%) and postoperative mortality (Additional file 2: Appendix 8) [1, 30, 32, 55, 56].

Table 2 Prospective studies of risk factors for postoperative mortality among older adults undergoing elective surgery
Full size table

Length of hospitalization

A total of 21 studies (5037 patients) investigated preoperative prognostic factors and length of hospitalization (Table 3) [1, 2, 7, 24, 25, 27, 29,30,31, 35,36,37, 39, 45, 46, 51, 53, 58,59,60,61]. Substantial between-study heterogeneity in the reporting of outcomes and few prognostic factors being reported in more than one study largely precluded pooling of study-level effect estimates. There was no association between higher American Society of Anesthesiologists (ASA) score and prolonged hospitalization (OR 0.82, 95% CI 0.30–2.23, 2 studies, I2 = 0%) (Additional file 2: Appendix 9) [2, 31]. Among six studies that investigated frailty as a prognostic factor for prolonged hospitalization, there was a significant association identified in four studies [7, 29, 35, 37, 40, 61]. One study found that, while frail patients with postoperative complications had prolonged hospitalizations (P < 0.001), those without postoperative complications did not (P = 0.19) [36]. Age was identified as a significant prognostic factor for prolonged hospitalization in only two of six studies [1, 2, 6, 26, 33, 51, 62,63,64].

Table 3 Prospective studies of risk factors associated with prolonged hospitalization among older adults undergoing elective surgery
Full size table

Destination at discharge from hospital

In total, 13 studies (2601 patients) investigated associations between preoperative prognostic factors and the destination at discharge from hospital (e.g., skilled nursing facility vs. discharge to home) (Table 4) [2, 7, 25, 29, 37, 38, 40, 45, 51, 53, 61, 65, 66]. Patients were discharged to a number of locations, including other hospitals, nursing homes, rehabilitation centres, transitional care facilities, and assisted-living facilities. The pooled incidence of discharge from hospital to a destination other than home was 13.65% (8.90–20.39%, 9 studies, I2 = 91.6%, NNF = 8) [2, 25, 37, 38, 40, 45, 51, 53, 66]. In a subgroup of older adults undergoing general surgery, the pooled incidence of being discharged to a non-home location was 9.97% (6.59–12.39%, 2 studies, I2 = 0%, NNF = 11) [2, 40]. Only the effects of publication year could be explored in a meta-regression because there were not enough studies to explore the effects of type of surgery or mean age of patients on the pooled incidence of mortality. Publication year did not explain any of the variance in the meta-regression model. In another subgroup of older adults, the pooled incidence of being discharged to a nursing home was 9.97% (5.30–17.96%, 2 studies, I2 = 86%) [37, 66]. Meta-analysis of data from five studies (1228 patients) found that frailty was associated with non-home discharge following elective surgery (OR 3.42, 95% CI 1.35–8.68, I2 = 67.46%) (Additional file 2: Appendix 10) [7, 29, 37, 38, 40]. In an additional study, the odds of being transferred to another hospital were six times greater for frail patients (P = 0.002) (Table 4) [61]. There were a number of prognostic factors that were associated with an increased risk of non-home destination at discharge from hospital, namely older age, weight loss ≥ 10%, ASA score ≥ 2, ECOG performance status ≥ 2, and lower self-reported mobility [2, 7, 37, 66, 67].

Table 4 Prospective studies of risk factors associated with non-home discharge among older adults undergoing elective surgery
Full size table

Functional decline

Six studies (1426 patients) investigated the association between preoperative prognostic factors and postoperative functional decline (Table 5) [63,64,65, 68,69,70]. All six studies reported prognostic factors associated with postoperative impairment in a patient’s ability to perform ADLs. One study reported risk factors associated with postoperative impairment in the ability to perform IADLs [63]. The pooled incidence of decline in ADLs was 21.03% (9.94–39.11%, 4 studies, I2 = 97.1%, NNF = 5) [63, 64, 69, 70]. In a subgroup of patients undergoing general surgery, the pooled incidence of decline in ADLs was 15.25% (5.48–35.83%, 2 studies, I2 = 95.7%, NNF = 7) [63, 64]. Age was not found to be associated with postoperative impairment in ADLs at 4–6 weeks, 3 months, or 1 year after elective surgery; however, one study did show an association between age and impairment in ADLs in the postoperative period [63,64,65, 68, 70]. Baseline MMSE score was not associated with a decline in ADLs, but it was associated with a decline in IADLs [63, 64].

Table 5 Prospective studies of risk factors associated with functional decline among older adults undergoing elective surgery
Full size table

Discussion

This systematic review and meta-analysis identified preoperative prognostic factors associated with the risk of harm in older adults undergoing elective surgery. Common geriatric syndromes, such as functional impairment, cognitive impairment, and frailty, were associated with the composite outcome of postoperative complications, while more traditional perioperative risk factors in the medical literature, such as older age and ASA status, were not [71]. Although the pooled incidences of adverse postoperative outcomes must be interpreted with caution because of significant between-study heterogeneity, it is worth noting that approximately one in four older adults suffered a postoperative complication from undergoing elective surgery. Fortunately, we identified a number of potentially modifiable risk factors, including smoking status, depressive symptoms, and frailty, that can be explored in future studies aimed at preventing adverse postoperative outcomes in older adults undergoing elective surgery.

The finding that geriatric syndromes, but not older age or ASA status, were associated with postoperative complications warrants further discussion. In particular, frailty is felt to represent a patient’s biological age as opposed to their chronological age, which may explain why frailty and not older age was associated with postoperative complications in this setting [72]. Frail patients were also less likely to be discharged to their home, which again likely reflects their decreased physiological reserve to respond to a significant stressor such as surgery. Besides being associated with postoperative complications, frailty has been associated with a number of other adverse outcomes outside of the perioperative literature, including mortality and admission to a long-term care facility [73, 74]. Perhaps greater emphasis should be placed on a patient’s frailty status as opposed to their age in determining risk of adverse postoperative outcomes as part of a comprehensive preoperative assessment [9].

The high incidence of adverse outcomes (25% of patients experiencing a postoperative complication), even in this non-emergent surgical setting, was also surprising. There was significant between-study heterogeneity among studies reporting postoperative complications, which could not be completely explained by type of surgery, but instead likely reflects the range of postoperative complications that were reported by study authors (e.g., atelectasis, venous thromboembolism, death). In the future, it will be important for more researchers to identify postoperative complications by severity so that knowledge users (e.g., patients, clinicians) can have a better-informed discussion as to a patient’s risk of developing different postoperative complications.

To our knowledge, this is the first systematic review and meta-analysis that comprehensively examined the association between preoperative prognostic factors and adverse postoperative outcomes among older adults undergoing elective surgery. A recent narrative review on adverse postoperative outcomes among older adults included patients with different indications for surgery, such as hip fracture or other emergent procedures, and did not conduct meta-analyses of prognostic factors [75]. We targeted older adults undergoing elective surgery because of the potential to intervene to improve patient outcomes by identifying and optimizing these factors preoperatively. Multicomponent interventions aimed at improving a patient’s nutrition, physical fitness, and cognition have shown promise in improving frailty [76]. Similarly, smoking status and depressive symptoms are potentially modifiable prognostic factors that were associated with developing postoperative complications. Interventions for preoperative smoking cessation have been associated with a lower risk of postoperative complications [77]. These prognostic factors could be targeted in the preoperative clinic.

There were limitations in our study’s review process. Firstly, only studies that were published in English were included in this review to increase feasibility, but our findings are likely generalizable given the number of geographical regions represented in our systematic review. Secondly, there was substantial heterogeneity between studies for some outcomes, which could not always be adequately explored given a limited number of studies and a lack of individual patient-level data. Indeed, it is possible that by including such a broad spectrum of elective surgical procedures we may create difficulty in understanding exactly which prognostic factors are most likely to be important for certain patients, but this was explored in subgroup analyses and meta-regression models, where possible. Additionally, this study was initiated prior to the introduction of the CHARMS checklist, which means that biases introduced in model development, validation, and evaluation of our included studies are less well described; however, we feel that we were able to identify important sources of selection bias, measurement bias, and confounding that threatened the validity of individual study findings [78].

There were also limitations imparted by the included studies themselves. The methodological quality assessment demonstrated that there were a number of studies reporting varying intensity of follow-up, which may have impacted the incidence of complications. The majority of studies included in this systematic review were cohort studies; therefore, our findings may be influenced by confounding. Sensitivity analyses demonstrated that our findings were largely consistent when only study-level effect estimates that were adjusted for potentially important confounders were included in the meta-analyses. Lastly, sometimes studies did not report independent variables for which there was a non-significant association with the dependent variable in the final multivariable model, which could potentially lead to a type 1 error in the findings of our meta-analyses. This is a limitation that is inherent in the prognosis literature that we hope will be overcome in the future by improved quality of reporting.

Our study had a number of strengths. There were 44 studies and over 12,000 patients included in our systematic review and meta-analyses, which allowed us to investigate a number of possible prognostic factors. The hypothesis-generating nature of this study allowed for the identification of prognostic factors that are potentially modifiable in the preoperative setting, which could lead to better surgical outcomes for older adults undergoing elective surgery.

Conclusions

In summary, this systematic review and meta-analysis highlights how common postoperative complications are among older adults undergoing elective surgery (NNF = 4) and the importance of geriatric syndromes in identifying older adults at risk of harm. Furthermore, there were several prognostic factors identified that could be modifiable in a preoperative setting, including smoking and frailty, which can be explored in future knowledge translation strategies to develop interventions aimed at mitigating the risk faced by older adults undergoing elective surgery.