Association Between Postoperative Long-Term Heart Rate Variability and Postoperative Delirium in Elderly Patients Undergoing Orthopedic Surgery: A Prospective Cohort Study

This prospective cohort study was conducted in the Zhujiang Hospital of Southern Medical University from June 1, 2019, to October 30, 2020. The study protocol was approved by the Institutional Committee for Medical Ethics (approval number 2020-KY-065-02, Medical Ethics Committee of Zhujiang Hospital of Southern Medical University, China). Patients were asked to give written informed consent to participate. We enrolled patients aged 60 or older who scheduled for orthopedic surgery [including total hip arthroplasty (THA), total knee arthroplasty (TKA), revision TKA (RTKA), revision THA (RTHA), hip fracture repair, and femoral shaft fracture (FSF) surgery] under spinal anesthesia. Exclusion criteria were as follows: (I) American Society of Anesthesiologists (ASA) grade ≥ IV; (II) clinical evidence of acute coronary artery disease in the last 3 months, significant valve disease; (III) arrhythmia (atrial fibrillation or flutter, ectopic beat > normal sinus 5% of pulsation); (IV) β-blocker user; (V) history of endocrine disease (thyroid or adrenal gland disease); (VI) Parkinson disease (PD), all-cause dementia (including PD-related dementia, AD-related dementia, and Lewy’s body dementia), CNS tumor, and stroke with hemiplegia; (VII) preoperative mini-mental state examination (MMSE) score <17 or preoperative delirium; (VIII) severe hearing impairment and inability to communicate; (IX) long-term usage of sedatives or other mental illness drugs, and drug or alcohol abuse; and (X) transferred to an intensive care unit (ICU) postoperatively.

With the end of the operation, participants were invited to wear a 12-lead Holter monitor (Biomedical Instruments Limited Company, Shenzhen, China) and then transferred back to quiet inpatient wards for 24 h. In order to prevent movement from interfering with HRV, during the recording process, participants were asked to relax as much as possible and lie in bed for a long time.

Holter monitor recordings were transmitted to the Dynamic ECG analysis system (Biomedical Instruments Limited Company, Shenzhen, China) and interpreted using the same software. HRV parameters were automatically documented as numerical data by the Holter software. Artifacts and arrhythmia were manually identified and removed by an experienced cardiologist blinded to the clinical data of the cases. Only the analyzable ECG data recorded for more than 18 h could be considered qualified.

Parameters in time domain and frequency domain were applied to HRV assessment. In the time domain, the following indices were calculated: SDNN (in milliseconds), SDNNI (in milliseconds), and RMSSD (in milliseconds). In the frequency domain, the power spectra of the following frequency bands were calculated: HR, HF (0.15–0.40 Hz, in normal units), LF (0.04–0.15 Hz, in normal units), VLF (0.0033–0.04 Hz, in milliseconds), ULF (≤0.003 Hz, in milliseconds), and TP (in milliseconds).

We interviewed all patients the day before the operation. Patients were screened for dementia through an MMSE test and were recruited with a score of more than 16. The assessment of delirium was conducted using the Chinese version of the 3-Min Diagnostic Interview for CAM-defined Delirium (3D-CAM) by investigators who were trained and supervised by a delirium expert. The evaluation included face-to-face interviews with patients and their accompanying family members or nursing assistants and review of the medical records and nurse’s notes since patients’ admission.

The 3D-CAM has four diagnostic features: (1) an acute change and fluctuating course, (2) inattention, (3) disorganized thinking, and (4) altered level of consciousness. Delirium is suggested if both features 1 and 2 are present at the same time that either feature 3 or 4 is identified. In addition, the 3D-CAM was chosen to detect POD due to its high sensitivity in comparison with the gold standard, the Diagnostic and Statistical Manual of Mental Disorders, 5th edition (DSM5), criteria (Marcantonio et al., 2014). Moreover, the Chinese version has also been proven to have good reliability and validity with use in the Chinese elderly population (Ji et al., 2020).

For delirium patients, the Chinese version of the Delirium Rating Scale-98-R (DRS-98-R) was additionally used to assess the subtype of delirium. According to the scale, delirium can be classified into three motoric subtypes. The hyperactive subtype was defined as a score of 1 to 3 on Item 7 of the DRS-98-R (motor agitation) and a score of 0 on Item 8 of the DRS-98-R (motor retardation). The hypoactive subtype was defined as a score of 0 on Item 7 of the DRS-98-R and a score of 1 to 3 on Item 8 of the DRS-98-R. The mixed subtype was defined as a score of 1 to 3 on Items 7 and 8 in the DRS-98-R or a score of 0 on both items (Huang et al., 2009).

Investigators performed delirium assessments 1 day before the scheduled surgery and daily up to the seventh postoperative day from 7 am to 9 am. Patients with POD were be followed up until discharged. Both the subtype and duration of delirium were required to be recorded.

Studies had shown that age, Charlson Comorbidity Index (CCI), and perioperative usage of benzodiazepine and opiates were associated with delirium after orthopedic surgery in elderly patients (Scott et al., 2015; Weinstein et al., 2018). In addition, body mass index (BMI), sex, and physical status were correlated with HRV (Shaffer and Ginsberg, 2017). Therefore, the following variables were collected as potential confounders: age, sex, BMI, CCI, surgery, dosage of midazolam, and equivalent morphine usage.

Demographic characteristics (weight, height, BMI, sex, and educational level), lifestyle habits (mainly drinking and tobacco smoking), and clinical features (test results, diagnosis, past medical history, type of surgery, surgery duration, and concomitant medications) were collected through a purpose-designed questionnaire, administered by an interviewer, and through available medical records.

A minimum of 10 events per variable are necessary to adequately produce estimates of effect with regression models (Concato et al., 1995). Based on the reported incidence of POD after orthopedic surgery of approximately 17.3% (Scott et al., 2015), a sample size of 289 individuals will allow five variables to be assessed in the regression model.

Descriptive statistics were expressed as the mean ± SD, median (interquartile range), or frequencies with percentages to compare characteristics of participants with and without POD including demographics, behavioral habits, and outcomes. HRV parameters were compared between the groups stratified by the presence of POD using Mann–Whitney U tests. Analyses further involved Spearman’s rank correlation coefficients of the entire cohort to evaluate a potential relationship between HRV parameters and POD. Then, Spearman’s rank correlation coefficients were used to estimate the associations between the HRV parameters and the duration of POD. Furthermore, HRV parameters that differed between the groups at the 20% level of significance (i.e., P < 0.20) were entered into multivariable logistic regression models for the outcome of POD. Covariates were chosen based on significant associations with the occurrence of POD and previous literature. All analyses were conducted using the Statistical Package for the Social Sciences (SPSS) version 25.

According to the exclusion criteria, 355 of the 451 patients were recruited between June 1, 2019, and October 30, 2020 (Figure 1). In the end, 294 patients completed follow-up. Sixty participants (20.4%) developed POD with a median duration of 1 day.

Characteristics of the cohort stratified by the presence of POD are shown in Table 1. Compared with patients not experiencing POD, delirium patients tended to be older and had a higher ASA PS, higher CCI, higher dosage of opiates, and higher percentage of hip surgery (including THA, RTHA, and hip fracture repair). They also had a lower preoperative hemoglobin (Hb) concentration and lower dosage of midazolam. In contrast, there were no differences in sex, education level, duration of surgery, drinking, and smoking between groups.

In patients with POD, 18 were assessed as hyperactive subtype of delirium, 17 as hypoactive subtype, and 25 as mixed subtype. Except for the duration of surgery, there was no difference in the characteristics among three subtypes. Compared with the other two subtypes, the operation time was shorter in mixed subtype (). Supplementary Table 1

Heart rate variability parameters of the cohort stratified by the presence of POD were expressed as median and reported in Table 2. Aside from RMSSD and HR, other HRV indices were higher in patients who experienced POD compared with those without delirium. These differences were significant for SDNNI (P = 0.027) and ULF (P = 0.006).

A comparison of HRV among three subtypes of delirium patients is illustrated in. No differences were observed in the HRV indices among groups. Supplementary Table 2

Table 3 shows the correlation analysis between HRV and POD. The occurrence of POD was significantly negatively correlated with SDNNI, VLF, ULF, and TP but not with SDNN, HF, or LF, which were traditionally of concern. As stratified into young-old (60–79 years old) and old-old (≥80 years old) participants by age, POD was still significantly negatively correlated with SDNNI, VLF, and ULF in young-old adults rather than in the old-old population (Supplementary Table 3).

Observed to be associated with cognitive function (Dalise et al., 2020), HF and LF were selected for logistic regression analyses. In addition, as SDNNI, VLF, and ULF were associated with POD at a significance level of less than 0.2, they were also included in the planned binary logistic regression analyses. After potential confounders were controlled for (i.e., sex, age, BMI, educational level, CCI, surgery, dosage of midazolam, and equivalent morphine), there was no significant association between HRV indices and POD (Table 4). Noticeably, after stratification by age, there was a significant association between ULF and POD in young-old adults (Table 5). Stratified by sex subsequently, SDNNI, VLF, and ULF were significantly associated with POD in male patients (Table 6).

The median duration of POD was 1 day. There were significant negative correlations of SDNNI, VLF, ULF, and TP but not RMSSD, HF, or LF with the duration of POD (Table 7).

A survey in an ICU observed that the delirium patients had a lower LF/HF ratio and a greater reduction in SDNN than the control group during head-up tilt to 15° (Neerland et al., 2019). In our results, the frequency and time domain parameters of patients with POD were mostly lower, which was consistent with the result of a research about HRV at mild cognitive impairment stage of Lewy body disease (Kim et al., 2018).

Conversely, a research in patients with hip fracture showed that preoperative SDNN, TP, and HF values were significantly higher in patients with delirium and that LF and LF/HF were lower compared with those in patients without delirium (Ernst et al., 2020). One reason for this discrepancy might be as follows: this investigation studied HRV by analyzing short-term measurements based on preoperative 5-min recordings, which could not accurately reflect environmental stimuli and the central nervous activity including circadian rhythms. Besides, at very low respiratory rates (<7–8 breaths per minute or deep breaths), parasympathetic activity can drive LF power and other indices. Therefore, respiratory rate must be considered, since it is a confounder of short-term measurements. It is also true that the LF/HF ratio has long been considered an index of sympathovagal balance, but this viewpoint has also been strongly criticized because the physiological bases are not clear (Billman, 2013; Shaffer and Ginsberg, 2017). What is more, the study focused on preoperative HRV indices, which changed so dramatically after surgery that they could not reflect the effects of operation and anesthesia exerted on patients.

As reported by Shaffer and Ginsberg (Shaffer and Ginsberg, 2017), the 24-h HRV recordings achieve greater predictive power than short-term measurements. Therefore, our study design provided 24-h ECG recordings postoperatively. Correspondingly, neither the levels of LF nor HF show a difference in our result. There is a significant correlation between POD and VLF, ULF, SDNNI, and TP power, represented as the gold standard for a more accurate evaluation of total circadian HRV (Singh et al., 2018a,b).

The study’s strengths include its prospective nature, blinding of assessors, and use of validated and reliable measures to assess delirium status and HRV parameters.

There are limitations to this work that should be noted. Primarily, the age span of participants included in the cohort is too large. According to the China country assessment report on aging and health, people over 60 are classified as the elderly (No-authors, 2016). Apparently, the young-old patients may not be representative of the entire elderly population. While life expectancy is rising, the health span has not kept up with it (Partridge et al., 2018). The aging process has been linked to the occurrence of chronic diseases and functional impairments, including frailty and metabolic and neurodegenerative diseases (Niccoli and Partridge, 2012; Rivero-Segura et al., 2020). Except for old-old population with hip fractures, a considerable part of the patients who underwent orthopedic surgery were young-old patients who suffered from arthritis. As a common chronic disease in the elderly, an European cross-sectional survey indicates that nearly 50% of elderly patients with symptomatic knee osteoarthritis show fragility (Salaffi et al., 2020). Particularly, fragility is not only a reflection of biologic rather than chronologic age but also an important predisposing factor of delirium (Bersani et al., 2020; Rivero-Segura et al., 2020; Susano et al., 2020).

Specially, due to disadvantageous education (Zhang et al., 2008), cognitive decline in elderly Chinese is more serious. As our research shows, only 25% of participants received high school education or above. Therefore, we set 60 years old as the lower age limit to prevent people who have a young calendar age but experiencing aging from being missed.

In addition, there is a wide spectrum of operative types in this cohort. The study included patients who suffered from hip fracture who were more likely to be older and have with more diseases. A well-known complication of hip fracture surgery is POD, with an incidence up to 56% (Oh et al., 2015). Therefore, we excluded patients with preoperative MMSE score < 17 and preoperative delirium who were more susceptible to POD. Besides, the corresponding baseline HRV of patients should be collected before surgery. Considering the influence of movement and anxiety on HRV before operation (Goessl et al., 2017; Hunt and Saengsuwan, 2018), we only record postoperative long-term HRV, which could reflect the state of patients after surgery more accurately. A more convincing conclusion may require further study to compare predictive power for the occurrences of POD upon HRV of patients before operation and post operation.

What is more, the development of delirium involves the complex inter-relationship between a vulnerable patient with multiple predisposing factors and exposure to noxious insults or precipitating factors (Inouye et al., 2014). Our assessment of the above factors is not comprehensive enough. Finally, due to the small sample size, we cannot adjust for other drugs used in the statistical analyses. Consequently, we excluded the patients who were transferred to the ICU for further treatment because of complex drug use and surroundings. In addition, although generally thought that there is no difference in the incidence of POD between general and regional anesthesia (Mason et al., 2010), we only included patients under spinal anesthesia to eliminate effects on HRV exerted by other anesthetics.

In this prospective cohort study, we investigated the association between postoperative long-term HRV parameters and POD in an orthopedic surgery elderly population. The results showed a negative correlation between VLF, ULF, SDNNI, and TP and incidence and duration of POD. This finding suggests that the occurrence of POD may affected by ANS change and circadian rhythm disorder reflected by postoperative long-term HRV parameters. Thus, future research on how ANS activity and circadian rhythm function affect POD in elderly patients would provide an opportunity to explain relevant pathophysiological factors in POD.