Predicting Postoperative Physiologic Decline After Surgery

This was a prospective analysis of 3295 patients undergoing preoperative evaluation for all scheduled all-condition elective surgical procedures and related procedures between May 1, 2017, and December 31, 2017, that required postoperative inpatient admission for recovery at a tertiary care freestanding children’s hospital. This preoperative evaluation was required to provide anesthetic evaluation and risk stratification for all elective surgeries requiring hospitalization. Surgical procedures performed by both general and specialty providers across the spectrum of technical complexity were included. This study was approved for human subjects research by the Boston Children’s Hospital and the Harvard T.H. Chan School of Public Health.

The main outcome measure was unanticipated, major PoPD that required immediate clinical attention. As part of routine clinical care, PoPD was assessed prospectively at least once every 4 hours for each patient by the assigned inpatient nurse, reviewed with the patients’ hospital team (eg, attending physician, resident, etc) for clinical action, and entered into the patient’s electronic health record by using standardized fields. PoPD assessments were integral to the hospital’s early warning system to identify and treat patients in physiologic distress.²³ Because physiologic decline may be expected and commonplace immediately after surgery, PoPD in the postanesthesia care unit and in the ICU was excluded from measurement.

PoPD was categorized by cardiovascular, respiratory, and neurologic symptoms. Cardiovascular decline included the following: (1) gray and/or mottled coloring; (2) capillary refill >4 seconds; (3) moderate or severe tachycardia; and (4) new-onset or worsening bradycardia, ectopy, irregular rhythm, or heart block. Respiratory decline included the following: (1) moderate or severe tachypnea; (2) hypopnea; (3) moderate or severe increased work of breathing; (3) >40% oxygen via face mask; (4) ≥1 L of oxygen via nasal cannula above baseline need; (5) nebulizer treatments ≤2 hours apart; (6) moderate to severe oxygen desaturations; and (7) apnea requiring stimulation, repositioning, or other interventions. Neurologic decline included the following: (1) increase in seizure frequency or duration, (2) new-onset focal neurologic deficit, and (3) insufficient responsiveness to painful stimulation. To evaluate instances of PoPD, 2 study team members (J.G.B. and I.L.) reviewed the electronic health record for vital signs, clinical notes, diagnostic studies, orders, and other information to ascertain the clinical circumstances and etiology of the PoPD (eg, tachycardia due to pain, hypovolemia, sepsis, etc) and the action taken to address the PoPD (eg, diagnostic studies, medications, transfer to higher level of care, etc).

All patients had a preoperative anesthesiology clinic visit performed within 30 days of the scheduled procedure, which consisted of an evaluation by an anesthesiologist-supervised registered nurse or nurse practitioner as well as an attending anesthesiologist. Active health problems were identified by this clinical team, using the definition that an issue was any health problem that could compromise the child’s perioperative health and safety. Active health issues were prospectively recorded with an electronic standardized form, organized by organ system. Examples of health problems included, but were not limited to, an acute illness (eg, an upper-respiratory illness or pneumonia) or an undiagnosed or uncontrolled high-severity chronic condition identified from review of systems (eg, obstructive sleep apnea).

Semistructured interviews and focus groups were conducted with perioperative clinical providers to set the process for recording active health issues in the electronic form. The form was field tested for 3 months to optimize feasibility and usability. Made mandatory through hospital policy, data were entered at the end of the preoperative clinic visit. Proactive monitoring occurred throughout the study period to ensure adherence with the data collection system.

Information on past medical history included home medications, American Society of Anesthesiologists’ (ASA) Physical Status Classification System score,²⁴ and type and number of chronic conditions. The number of chronic medications taken by each child at home at the time of the preoperative evaluation was documented. The number of chronic medications was recorded in categories (0, 1–5, 6–10, and ≥11). Complex chronic conditions (CCCs) were assessed using Feudtner’s pediatric CCCs version 2 International Classification of Diseases, Ninth Revision, Clinical Modification codes.^25,–28 The CCCs are life-limiting childhood health conditions that are associated with severe limitations in function as well as high morbidity and mortality.^3,25,–31 

The characteristics included age, sex, race and ethnicity (non-Hispanic white, non-Hispanic African American, Hispanic, other), and home residence (in state, out of state, and international).

Because of the known and expected variation in physiologic stress and postoperative recovery across different types of elective surgical procedures, PoPD was aggregated by surgical procedure. The 266 unique surgical procedures were ranked by the percentage of patients experiencing PoPD and then categorized into lower and higher risk groups at the binary split of ≥20%.

We used χ² testing to assess the associations between categorical patient characteristics and PoPD. Multivariable logistic regression models were derived with a focus on active health problems identified during the preoperative evaluation as well as polypharmacy to obtain odds ratios, 95% confidence intervals (CIs), and P values assessing the independent associations between patient characteristics and PoPD, while controlling for known confounding variables (eg, chronic conditions and type of surgery). There were no missing data for the variables or outcome. Model performance was assessed with the concordance statistic and the Hosmer-Lemeshow goodness-of-fit test. Classification and regression tree (CART) analysis was used to assess which combinations of patient characteristics had the strongest associations with PoPD. Analyses were conducted by using R (version 3.4.3; R Foundation for Statistical Computing, Vienna, Austria) and Stata (version 14; StataCorp, College Station, TX). A two-sided α level was set at P < .05.

Of the 3295 patients, median age was 8 years (interquartile range [IQR]: 2–15 years). Of the 12.9% of patients who were 18 years or older, most were young adults (median: 20 years; IQR: 18–23 years). Regarding other demographics, 1671 (50.7%) were male, and 1844 (56.0%) were non-Hispanic white (Supplemental Fig 5). Otolaryngology (23.7%), orthopedics (25.1%), and general surgery (16.5%) were the most common surgical subspecialty categories. Tonsillectomy or adenoidectomy (14.3%), spinal fusion (7.2%), and gastrostomy (3.1%) were among the most common surgical procedures (Supplemental Table 2).

Seventy-one percent of patients (n = 2336) had a CCC; congenital or genetic CCCs (33.4%) and cardiovascular CCC (26.5%) were the most common. Twenty-eight percent of children (n = 933) were assisted with medical technology. Sixty-four percent (n = 2118) used 1 or more chronic medications, 44.5% (n = 1467) used 1 to 5 medications, 12.4% (n = 410) used 6 to 10 medications, and 7.3% (n = 241) used ≥11 medications. Thirty-seven percent (n = 1226) of children had an ASA score ≥III (severe systemic disease).

The preoperative evaluation occurred a median of 9 (IQR: 4–15) days in advance of the scheduled surgery. Forty-seven percent of patients (n = 1556) had an active health issue. Nine percent (n = 282) had an acute illness; 4.3% (n = 142) had an abnormal preoperative laboratory or radiographic test; 14.1% (n = 463) had a concern regarding family history or a previous anesthesia-related complication, risk factor, or incident; and 11.2% (n = 368) required attention (eg, surgical clearance) related to assistive technology (eg, cerebrospinal fluid ventricular shunt).

PoPD occurred in 26.8% of patients (n = 882) (Table 1). The percentages of patients with cardiovascular, respiratory, and neurologic PoPD were 19.4% (n = 641), 11.7% (n = 386), and 3.4% (n = 111), respectively. The median number of hours from surgery to PoPD was 34 (IQR: 16–94). Among the most common etiologies for PoPD were hypovolemia; respiratory compromise due to pulmonary secretions, atelectasis, obstruction, and pneumonia; and fever and sepsis. The most common clinical action and treatments to address PoPD were intravenous fluids, enhanced respiratory support (eg, oxygen, recurrent suctioning, nebulizers), antibiotic administration, and clinical consultation (eg, with specialists and ICU providers) for enhanced assistance. No patients died during their hospital stay.

In bivariable analysis, a higher percentage of patients with versus without an active health issue during preoperative evaluation experienced PoPD (31.7% vs 22.4%, P < .001) (Fig 1). A higher percentage of patients with versus without the following specific types of active health problems experienced PoPD: oncologic (55.0% vs 26.0%, P < .001), neurologic (43.2% vs 24.9%, P < .001), nephrologic (41.4% vs 26.3%), and respiratory (33.9% vs 26.1%, P < .001) (Fig 1).

In bivariable analysis, a higher percentage of patients using 11 or more versus 10 or fewer chronic medications experienced PoPD (44.0% vs 25.4%, P < .001) (Fig 2). As the ASA score increased from I to IV, the percentage of patients with PoPD increased from 16.1% to 44.1%, P < .001. As the number of CCCs increased from none to 3 or more, the percentage of patients with PoPD increased from 13.0% to 37.7% (P < .001) (Fig 2). The highest percentages of PoPD were observed with versus without the following CCCs: malignancy (43.8% vs 24.2%, P < .001), metabolic (39.9% vs 25.1%, P < .001), neuromuscular (39.5% vs 25.1%, P < .001), and hematology or immunology (38.5% vs 25.3%, P < .001) (Fig 3).

As expected, the percentage of patients with PoPD varied significantly (P < .001) across types of surgery (Supplemental Table 2). Examples of surgical procedures with the lowest percentages (0%–9%) of PoPD were choanal atresia repair and supraglottoplasty. Examples of surgical procedures with the highest percentages of PoPD (≥40%) were hemispherectomy and spinal fusion (Supplemental Table 2).

In multivariable logistic regression, the adjusted odds ratios of PoPD were 1.2 (95% CI: 1.0–1.4) for presence versus absence of an active health problem at the preoperative clinical evaluation; 1.6 (95% CI: 1.3–2.1) for the presence of 1 CCC, 1.7 (95% CI: 1.3–2.3) for the presence of 2 CCCs, and 2.2 (95% CI: 1.7–2.9) for the presence of ≥3 CCCs, as compared with a reference group with no CCCs present (Fig 4). For the number of home medications, the adjusted odds ratios of PoPD were 0.9 (95% CI: 0.7–1.1) for 1 to 5 medications, 0.9 (95% CI: 0.7–1.1) for 6 to 10 medications, and 1.4 (95% CI: 1.0–1.9) for ≥11 medications, as compared with a reference group with no home medications. The adjusted odds ratio for higher- versus lower-risk procedures was 4.1 (95% CI: 3.4–5.0). In CART analysis, the lowest rate of PoPD (8.6%) occurred in children without an active health problem at the preoperative clinical evaluation; the highest rate (57.2%) occurred in children with at least 1 CCC who used ≥11 home medications (Fig 4).

In the current study, PoPD occurred in 1 in 4 pediatric patients undergoing elective surgical procedures that required admission for recovery in a freestanding children’s hospital. During preoperative evaluation, those patients had a high rate of active medical issues that required clinical attention to optimize their health and safety. PoPD risk was determined by the interactions of those active health issues with the specific surgical procedure, number of CCCs, and chronic medications. Combinations of those characteristics distinguished the patients with the lowest and highest likelihood of PoPD.

To our knowledge, the current study is the first to report the rates, types, and impact of health issues from all etiologies identified before elective surgical procedures in pediatric patients. The multivariable regression findings from the current study suggest that having an active health issue detected during preoperative evaluation contributes to the prediction of risk for PoPD, even after adjustment for chronic conditions, chronic medications, and type of surgery. Although this finding may seem intuitive, it is important to convey because the preoperative active health issues in the current study are not routinely included in existing tools that are used to assess surgical risk.^12,–20 Rather, those tools emphasize the presence of a chronic condition in a specific organ system (eg, respiratory) without incorporating perioperative clinical judgment on the effects of that condition and its impact on the child’s perioperative health and safety. In the CART analysis of the current study, it was revealed that preoperative active health issues had the strongest influence on PoPD with lower-risk procedures (eg, tonsillectomy or adenoidectomy); for those procedures, PoPD was nearly 80% more likely in patients with versus without a preoperative active health issue. It is imperative that pediatricians consider these findings because they can be used to help address preoperative active health issues and guide both surgical and anesthesiology providers when determining whether it is safe to proceed with surgery.^5,–9 

Beyond active health issues identified preoperatively, the current study is also the first to quantify the interaction of CCCs and polypharmacy with the likelihood of PoPD. In CART analysis, this interaction was the strongest predictor of PoPD for pediatric patients undergoing higher-risk surgeries (eg, spinal fusion for neuromuscular scoliosis); the likelihood of PoPD increased by over one-third in patients with a CCC when they used 11 or more chronic medications. This new finding in the pediatric population is congruent with emerging investigations of frailty and surgical risk (as manifested through chronic conditions and polypharmacy) in the elderly, adult population.^32,–34 In elderly persons, the stress of surgery can disrupt physiologic homeostasis, especially when there are preoperative baseline impairments in functional reserve across multiple organ systems. It is interesting that a comparable population of pediatric patients undergoing surgery exists. Although it may seem intuitive that greater frailty is associated with worse postoperative outcomes, the attributes of frailty in adult patients are increasingly used in real-time clinical decision support with precision medicine initiatives to estimate risk for individual surgical patients. Consideration of testing, adaptation, and use of frailty and related measurements used by general medical providers in the elderly population (eg, internists and geriatricians) may be warranted for use by pediatric providers in children to more accurately estimate and project postoperative outcomes.³⁵ 

The percentages of pediatric patients experiencing PoPD are higher than that reported in most previous studies. In those studies, including national quality improvement initiatives in the pediatric population that focus on specific postoperative complications (eg, surgical site infection), authors report a lower rate of them.^{14,18,–20,36} When compared with those previous initiatives, the methods used in the current study measured PoPD with a broader range of etiologies and severity. That may explain why our findings are comparable with the few prospective surgical studies that focus on more general organ system problems (eg, postoperative respiratory “problems”).^37,38 Moreover, PoPD assessment in the current study occurred at least 6 times daily by inpatient nurses. That method may have been more sensitive in the early detection of postoperative health issues that required clinical attention. Further investigation of risk factors for different gradations in the severity and types of physiologic deterioration associated with PoPD is needed.

Although in the current study we were not positioned to directly assess the relationships among PoPD, quality of care, and patient harm, future studies should assess which instances PoPD were preventable with high-quality perioperative care, especially those unexpected instances that resulted in significant adverse consequences. In this regard, there may be great value in the complementary use of predictive data in which both PoPD endpoints and actual adverse data are used, such as that reported in the National Surgical Quality Improvement Program.¹⁴ A first step of that data integration could entail identifying which PoPD measures (alone and in combination) are most predictive of the serious adverse events. Subsequent analysis of patient characteristics and preoperative measures of disease activity associated with both adverse events and clinically relevant PoPD profiles could then be used to more effectively identify high-risk patients during preoperative evaluation. Such an approach may provide a more comprehensive assessment to identify patients who are not only at risk for serious adverse events but for the PoPD profiles that are most predictive of these events.

This study has several caveats and limitations. In-person clinic visits were used for preoperative evaluations. Our institution uses this approach to adhere with ASA guidelines.^10,11 Across institutions, there is variation in the approach used for preoperative evaluation of elective surgical procedures (eg, chart review only). Further investigation is needed to assess which active health issues are identified through a different approach and how the issues affect the likelihood of PoPD. Although the current work is from a single-center tertiary care institution, the study methods and findings may be generalizable to other hospitals, especially those that currently employ or are developing early warning systems to detect physiologic decline in hospitalized patients^39,–43 and those that are participating in quality improvement initiatives to standardize patients’ preoperative evaluations performed by anesthesiology providers.⁴⁴ Moreover, one-half of the patients in our cohort had 0 or only 1 chronic condition; the PoPD findings in those patients may be helpful for hospitals that predominately perform surgeries and related procedures on patients who are less complex. The study was not positioned for us to assess all patient attributes (eg, functional limitations) that might also influence the likelihood of PoPD. Despite the use of the number of chronic conditions and medications in analysis, the true severity and fragility of the patients’ underlying health and well-being could not be quantified.

Wide-ranging rates of PoPD for pediatric patients undergoing elective surgery were revealed in the current study. Combinations of active health issues identified preoperatively along with the type of elective surgery, number of CCCs, and number of chronic medications distinguished the patients with the lowest and highest likelihood of PoPD. Consideration of those combinations in surgical risk tools may help optimize their use by perioperative providers during discussions of risks, benefits, informed consent, projection of outcomes, and decisions on whether to proceed with elective surgery. In addition, the study findings could lead postoperative hospital providers to be more vigilant about PoPD in children with the highest risk for it. This may be particularly helpful to promptly detect and address evolving postoperative hypovolemia, respiratory compromise, and other potentially avoidable events in children with medical complexity before such problems result in significant physiologic distress. Relatedly, the study findings may be useful when making nurse-patient assignments (eg, nurse-to-patient staffing ratio) to ensure that the patients at the highest risk of PoPD receive sufficient clinical attention. All of these clinical care considerations may be used to help inform subsequent investigations of integrated perioperative care across disciplines that strive to optimize the health and safety of children undergoing elective surgery.

We thank Mary Gibson, MSN, RN, Mary Pennelli, BSN, RN, Gretchen Doonan, BSN, RN, and Maureen Gormley, MSN, RN, CPNP for their time and effort to assist with data collection for this work and to prepare children with medical complexity for surgery.

ASA

American Society of Anesthesiologists

CART

classification and regression tree

CCC

complex chronic condition

CI

confidence interval

IQR

interquartile range

PoPD

postoperative physiologic decline