Preoperative predictors of acute postoperative anxiety and depression using ecological momentary assessments: a secondary analysis of a single-centre prospective observational study

Abstract

Background

Postoperative anxiety and depression can negatively affect surgical outcomes and patient wellbeing. This study aimed to quantify the incidence of postoperative worsening anxiety and depression symptoms and to identify preoperative predictors of these conditions.

Methods

This prospective, observational cohort study included 1168 patients undergoing surgery lasting >1 h with overnight admission at a university-affiliated quaternary referral centre. Postoperative anxiety and depression symptoms were measured using standardised, thrice-daily ecological momentary assessments (EMAs) for 30 days. Co-primary outcomes were worsening anxiety and depression symptoms, each defined as a slope >0 when EMA was modelled as a linear function of time. Multivariable logistic regression was performed to identify independent preoperative predictors of each outcome.

Results

Postoperative worsening anxiety occurred in 60 patients (5%), and postoperative worsening depression occurred in 86 patients (7%). Predictors of postoperative worsening of anxiety symptoms included preoperative Patient-Reported Outcome Measurement Information System (PROMIS) anxiety symptoms (adjusted odds ratio [aOR] 2.48, 95% credible interval [CI] 1.29–4.79, for mild symptoms; aOR 2.22, 95% CI 1.10–4.51, for moderate to severe symptoms), and preoperative pain (aOR 3.46, 95% CI 1.32–9.12). Predictors of postoperative worsening depression symptoms included preoperative PROMIS depression symptoms (aOR 2.26, 95% CI 1.24–4.14, for mild symptoms; aOR 3.79, 95% CI 2.10–6.81, for moderate to severe symptoms). Self-reported history of anxiety or depression did not independently predict either outcome.

Conclusions

Postoperative worsening anxiety and depression appear to be associated more closely with preoperative active mental health or pain symptoms rather than self-reported history of these conditions. Preoperative identification of at-risk patients will require screening for symptoms rather than simple history taking.

Editor's key points.

• •Postoperative anxiety and depression can negatively affect surgical outcomes.
• •The authors quantified the incidence of postoperative anxiety and depression symptoms to identify preoperative predictive factors for patients undergoing major noncardiac surgery.
• •There was a 5% incidence of worsening postoperative anxiety symptoms and a 7% incidence of worsening postoperative depression symptoms during 30 days after major surgery.
• •Severity of preoperative anxiety and depression symptoms were important predictors.
• •Screening patients for anxiety and depression before and after surgery could facilitate tailored interventions to improve postoperative outcomes.

Depression and anxiety are common psychological responses experienced by individuals following surgery.¹ Often, these responses coexist, further complicating patient recovery and overall wellbeing.² As with many psychiatric and medical diagnoses, subclinical levels of anxiety and depression that do not meet diagnostic criteria are common and also have important implications. Existing psychiatric comorbidities such as anxiety and depression have been associated with increased surgical mortality rates, more complications, longer hospital stays, higher readmission rates, impaired wound healing, and diminished overall quality of life.3, 4, 5, 6, 7 Identifying patients who are at risk for new onset or worsening postoperative anxiety and depression could assist clinicians in implementing interventions to prevent or mitigate adverse postoperative effects.

Previous studies have primarily used screening tools such as the Hospital and Anxiety Depression Scale and the State-Trait Anxiety Inventory to assess postoperative depression and anxiety. However, these assessment tools are administered at discrete time points such as 6 weeks, 6 months, and 1 yr, which miss clinically important symptoms that occur between assessments.8, 9, 10, 11 In addition, prior studies have typically examined narrow segments of the population defined by specific surgical subspecialties, with many segments of the broader surgical population unstudied.^4,7,8 Therefore, we aimed to characterise the incidence and predictors of postoperative worsening anxiety and depression symptoms, captured with high temporal resolution using ecological momentary assessment (EMA) in a broad surgical population. EMA involves real-time data collection from individuals in their natural environment to capture momentary experiences, behaviours, and environmental factors, providing a comprehensive understanding of symptom dynamics.¹² Based on previous literature, we hypothesised that older age, female sex, psychiatric comorbidity, and higher preoperative pain levels would be significantly associated with postoperative worsening anxiety and depression as measured by EMA.

Methods

This is a secondary analysis of the Personalized Prediction of Persistent Postsurgical Pain (P5) study (ClinicalTrials.gov identifier: NCT04864275). P5 is a single-centre, prospective observational cohort study with the goal of identifying predictors of persistent postsurgical pain. Subjects provided written informed consent. This project was approved by the Washington University Institutional Review Board (approval #202101123). P5 is an ongoing study, and this retrospective analysis included participants who had surgery between the time of study initiation on September 30, 2021, and the time of the data query on June 17, 2023.

Participants and setting

This study was conducted at Barnes-Jewish Hospital, a university-affiliated quaternary referral centre in St. Louis, MO, USA. Inclusion criteria for P5 included patients between the ages of 18–75 yr who were undergoing major surgery and had access to a smartphone (which was necessary for the EMA). Major surgery was defined as a procedure with an expected duration of >1 h and an anticipated overnight admission to the hospital or any of the following procedures even if not admitted: total joint replacement, hysterectomy, inguinal hernia repair, breast surgery, and video-assisted thoracic surgery. Patients who did not speak English were excluded from the study. Surgical candidates matching the inclusion criteria were identified using daily electronic health record-based reports sent to study coordinators. Information about the study was provided to eligible surgery candidates via an electronic health record patient portal or a telephone call from a study team member. Interested patients were consented primarily at the Washington University Center for Preoperative Assessment and Planning.

Data source and measurements

Standard of care was not affected for any of the participants because of study participation. Preoperatively, patients completed surveys including the Patient-Reported Outcomes Measurement Information System (PROMIS) anxiety and depression scales (short form),^13,14 pain catastrophising scale,¹⁵ and additional questions about preoperative pain, analgesic use, and past medical conditions. Use of the PROMIS anxiety and depression scales to measure anxiety and depression has been previously validated and used in clinical studies to assess perioperative mental health.^7,13,14,16 PROMIS anxiety and depression raw scores were converted to t-scores which were then categorised into ‘none to slight’ (t-score <55.0), ‘mild’ (t-score 55.0–59.9), ‘moderate’ (t-score 60.0–69.9), or ‘severe’ (t-score ≥70.0) anxiety or depression.¹⁷

After surgery, participants received smartphone notifications to complete EMA ratings thrice daily according to one of three schedules (07:00/12:00/17:00, 08:00/13:00/18:00, or 09:00/14:00/19:00). The participant could pick their preference from among these three schedules and had 60 min to complete the EMA after each notification. EMA continued through 30 days after surgery. Because it would not be feasible for participants to complete the entire seven-item PROMIS anxiety scale and seven-item PROMIS depression scale at every time point, the EMA included three items from each scale that had the highest item–total correlations with the parent scale in a previous cohort¹⁸ and other items related to pain. The six EMA prompts included in this study are shown in Supplementary Table S1, and participants responded to each question with a score between 0 (none) and 100 (worst possible).

Patient characteristics including age, sex assigned at birth, race, ethnicity, BMI, and medical history of anxiety, depression, diabetes mellitus, heart disease, hypertension, and cancer were obtained from patient reports. Preoperative medication usage, including opioids, benzodiazepines, selective serotonin reuptake inhibitors (SSRIs), and serotonin–norepinephrine reuptake inhibitors (SNRIs), was also obtained from patient reports. The scheduled surgery was retrieved from the operating room schedule, and surgeries were classified through the detection of keywords in the name of the surgery (Supplementary Table S2).

Sample size

This secondary analysis of the P5 study used a convenience sample of patients who had been enrolled in the parent study. No separate sample size calculation was performed, as the expected incidence of the primary outcomes was unknown.

Statistical methods

The co-primary outcomes were postoperative worsening EMA anxiety and postoperative worsening EMA depression. Dynamic structural equation modelling was used to examine person-level changes in depression and anxiety EMA ratings over the first 30 days postoperatively. A linear relationship was modelled between EMA ratings and time, with both random slopes and random intercepts at the patient level. Patients were classified as having postoperative worsening EMA anxiety or depression if they had a positive slope and the 95% credible interval (CI) did not contain 0 (Fig. 1).¹⁹

Fig 1.

Illustration of postoperative Ecological Momentary Assessment (EMA) outcomes. EMA was modelled as a linear function of time, with random slopes and intercepts for each participant. Participants were classified based on their random slope as experiencing either postoperative worsening (slope >0 and 95% credible interval [CI] excludes 0—purple line), postoperative improving (slope <0 and 95% CI excludes 0—goldenrod line), or no significant change (95% CI includes 0—green line). Participants with persistent severe symptoms (intercept in top decile, 95% CI for slope includes 0—blue line) were not classified as experiencing the outcome for the primary analysis but were reclassified as experiencing the outcome for the sensitivity analysis.

Categorical variables are presented with frequencies followed by percentages, and continuous variables are presented with means and standard deviations for normally distributed variables and median and interquartile range (IQR) for non-normally distributed variables. Comparisons between patients experiencing postoperative worsening EMA anxiety or depression and patients not experiencing postoperative worsening were performed using t-tests for normally distributed continuous variables, the Mann–Whitney U-test for non-normally distributed continuous variables, and χ² analysis for categorical variables. A P-value <0.05 was considered statistically significant. Following univariate comparisons, multivariable logistic regression was performed to identify independent predictors of worsening postoperative EMA anxiety or depression while adjusting for potential confounding variables identified using directed acyclic graphs (Supplementary Figs S1 and S2). Missing values for predictor variables occurred in 2% of patients and were imputed using multiple imputations with chained equations (with five imputations). This approach assumes that data were missing at random, which is particularly plausible here given the small amount of missing data. Dynamic structural equation modelling was performed using Mplus version 8.9 (Muthen & Muthen, Los Angeles, CA, USA).²⁰ All other analyses were conducted using R version 4.3.0 (R Foundation for Statistical Computing, Vienna, Austria) (2023-04-21).

Sensitivity analyses

Because the co-primary outcomes were defined using only the slope of postoperative EMA ratings, patients who reported persistent severe symptoms (high intercept, zero slope) would not be identified. Therefore, we conducted a sensitivity analysis where the outcome was a composite of postoperative worsening EMA anxiety or depression (as defined above) and persistent severe EMA anxiety or depression (with an EMA intercept in the top decile of the study population and an EMA slope 95% CI that includes 0; Fig. 1).

Some patients might have an EMA slope that is statistically >0 but too low in magnitude to be clinically relevant. Unfortunately, there is no prior literature to define the minimal clinically important difference in EMA measurements. To account for this, we conducted a second sensitivity analysis where patients whose random slopes fell in the bottom 5% were reclassified as having no postoperative worsening.

Results

Participants and follow-up

We included 1168 patients in this analysis (Fig. 2). Table 1 provides demographic and clinical information of the participants. Preoperative assessment, including baseline PROMIS surveys, occurred a median of 11 (IQR 6–17) days before surgery. Of the included patients, 470 (40%) self-reported a history of anxiety, and 427 (37%) self-reported a history of depression. The preoperative PROMIS anxiety survey revealed none to slight symptoms in 690 (59%), mild symptoms in 228 (20%), moderate symptoms in 214 (18%), and severe symptoms in 30 (3%). The preoperative PROMIS depression survey revealed none to slight symptoms in 849 (73%), mild symptoms in 170 (15%), moderate symptoms in 127 (11%), and severe symptoms in 16 (1%). There were 27 patients with missing values for one or more variables.

Fig 2.

Patient flow diagram. EMA, Ecological Momentary Assessment.

Table 1.

Patient characteristics (N=1168). PROMIS, Patient-Reported Outcomes Measurement Information System. ∗Scores range from 0 to 52, where higher scores indicate worse outcomes. ^†t-Scores converted to categorical variables: ‘none to slight’ (t-score <55.0), ‘mild’ (t-score 55.0–59.9), ‘moderate’ (t-score 60.0–69.9), or ‘severe’ (t-score ≥70.0).

Variable	Values	Missing values (n)
Sex, n (%)		0
Male	336 (29)
Female	832 (71)
Age (yr), median (interquartile range)	53 (42–64)	0
BMI (kg m−2), median (interquartile range)	31.7 (26.8–37.7)	0
Ethnicity, n (%)		0
Hispanic	22 (2)
Non-Hispanic	1146 (98)
Race, n (%)		0
American Indian or Alaskan Native	7 (1)
Asian	7 (1)
Black	207 (18)
Native Hawaiian or Other Pacific Islander	1 (<1)
White	919 (79)
Other	15 (1)
Multiracial	12 (1)
Type of surgery, n (%)		0
Abdominal	586 (50)
Breast	34 (3)
Head and neck	77 (7)
Orthopaedic	208 (18)
Spine	116 (10)
Thoracic	78 (7)
Vascular	27 (2)
Other	42 (4)
Diabetes mellitus, n (%)	189 (16)	0
Hypertension, n (%)	511 (44)	0
Heart disease, n (%)	91 (8)	0
Cancer, n (%)	346 (30)	0
Selective serotonin reuptake inhibitor, n (%)	204 (17)	0
Serotonin–norepinephrine reuptake inhibitor, n (%)	143 (12)	0
Benzodiazepine, n (%)	146 (12)	0
Opioid, n (%)	231 (20)	0
Self-reported history of anxiety, n (%)	470 (40)	5
Self-reported history of depression, n (%)	427 (37)	10
Self-reported history of both anxiety and depression, n (%)	334 (29)	5
Chronic pain, n (%)	333 (29)	11
Preoperative current pain, n (%)		6
No	287 (25)
Yes, and related to my need for surgery	581 (50)
Yes, and not related to my need for surgery	294 (25)
Pain Catastrophising Score,∗ median (interquartile range)	8 (3–18)	3
Preoperative PROMIS depression†, n (%)		6
None to slight	849 (73)
Mild	170 (15)
Moderate	127 (11)
Severe	16 (1)
Preoperative PROMIS anxiety†, n (%)		6
None to slight	690 (59)
Mild	228 (20)
Moderate	214 (18)
Severe	30 (3)

The EMA data collection period spanned from postoperative day 1 to day 30 for all patients. The median number of total EMA responses a patient completed was 64 (IQR 37–79). Participants completed EMA surveys a median of 28 total days (IQR 19–30) out of the 30-day study period, indicating active participation throughout the study duration (Supplementary Fig. S3). Responses to individual EMA items were highly correlated with responses to other items (Supplementary Fig. S4). Most participants demonstrated improvement in symptoms over time, and those with more severe preoperative symptoms had more severe postoperative symptoms (Fig. 3).

Fig 3.

EMA values over time for (a–c) anxiety and (d–f) depression. Patients were stratified by presence/absence of self-reported history of anxiety/depression and by the severity of preoperative PROMIS anxiety/depression symptoms. Data points represent the median EMA response in each stratum for all patients who responded at that time point. Error bars represent the interquartile range. (a) Patients with preoperative PROMIS anxiety t-score ≥70.0 with (n=180) and without (n=64) self-reported history of anxiety. (b) Patients with preoperative PROMIS anxiety t-score 60–69.9 with (n=101) and without (n=127) self-reported history of anxiety. (c) Patients with preoperative PROMIS anxiety t-score <60.0 with (n=188) and without (n=502) self-reported history of anxiety. (d) Patients with preoperative depression t-score ≥70.0 with (n=99) and without (n=43) self-reported history of depression. (e) Patients with preoperative depression t-score 60.0–69.9 with (n=95) and without (n=74) self-reported history of depression. (f) Patients with preoperative depression t-score <60.0 with (n=232) and without (n=614) self-reported history of depression. EMA, Ecological Momentary Assessment; PROMIS, Patient-Reported Outcomes Measurement Information System.

Postoperative worsening of anxiety

Of 1166 patients (excluding two patients who had depression EMA data but no anxiety EMA data), the EMAs revealed postoperative worsening EMA anxiety in 60 (5%), no significant change in 686 (59%), and significant improvement in 420 (36%). Univariate predictors of postoperative worsening EMA anxiety included younger age, preoperative pain (compared with no preoperative pain), and higher preoperative PROMIS anxiety (Table 2). In the adjusted multivariable logistic regression model, significant predictors of postoperative worsening EMA anxiety included mild preoperative PROMIS anxiety symptoms (aOR 2.48, 95% CI 1.29–4.79, P=0.007), moderative preoperative PROMIS anxiety symptoms (aOR 2.22, 95% CI 1.10–4.51, P=0.03), and preoperative pain (aOR 3.46, 95% CI 1.32–9.12, P=0.001). Self-reported history of anxiety was not a significant predictor of postoperative worsening EMA anxiety after adjusting for preoperative PROMIS anxiety. Results were similar in sensitivity analyses reclassifying patients with persistent severe anxiety as experiencing the outcome (Supplementary Table S2) and reclassifying 5% of patients with the smallest increases in postoperative anxiety as not experiencing the outcome (Supplementary Table S3).

Table 2.

Predictors of postoperative worsening anxiety EMA symptoms. CI, credible interval; EMA, Ecological Momentary Assessment; PROMIS, Patient-Reported Outcomes Measurement Information System. ∗Adjusted for all other variables shown in the table. The multivariable logistic regression model had a C-statistic of 0.71 and demonstrated a good fit using the Hosmer–Lemeshow test: χ²(8)=5.86, P=0.66. ^†t-Scores converted to categorical variables: ‘none to slight’ (t-score <55), ‘mild’ (t-score 55.0–59.9), ‘moderate’ (t-score 60.0–69.9), or ‘severe’ (t-score ≥70).

Variable	With worsening anxiety (n=60)	Without worsening anxiety (n=1106)	Unadjusted odds ratio (95% CI)	Unadjusted P-value	Adjusted odds ratio∗ (95% CI)	Adjusted P-value
Sex, n (%)
Female	39 (65)	791 (72)	(reference)		(reference)
Male	21 (35)	315 (28)	1.35(0.78–2.33)	0.28	1.63 (0.89–2.98)	0.12
Age (yr), median (IQR)	50 (39–60)	53 (42–64)	0.98(0.96–0.99)	0.03	0.98 (0.96–1.00)	0.05
Type of surgery, n (%)
Abdominal	29 (48)	556 (50)	(reference)		(reference)
Head, neck, and spine	16 (27)	176 (16)	1.74 (0.92–3.28)	0.09	1.36 (0.69–2.66)	0.38
Orthopaedic	10 (17)	198 (18)	0.97 (0.46–2.02)	0.93	0.91 (0.41–2.04)	0.83
Other	5 (8)	176 (16)	0.54 (0.21–1.42)	0.22	0.47 (0.17–1.27)	0.14
Self-reported history of anxiety, n (%)	27 (45)	443 (40)	1.21 (0.72–2.05)	0.46	0.84 (0.46–1.54)	0.58
Chronic pain, n (%)	19 (32)	314 (29)	1.15 (0.66–2.02)	0.62	0.84 (0.46–1.53)	0.58
Preoperative current pain, n (%)	55 (92)	819 (74)	3.77 (1.50–9.52)	0.005	3.50 (1.33–9.21)	0.01
Preoperative PROMIS anxiety†, n (%)
None to slight	23 (38)	667 (61)	(reference)		(reference)
Mild	18 (30)	208 (19)	2.51 (1.33–4.74)	0.005	2.46 (1.28–4.74)	0.007
Moderate to severe	19 (32)	225 (20)	2.45 (1.31–4.58)	0.005	2.22 (1.10–4.49)	0.03

Postoperative worsening of depression

Of 1168 patients, the EMAs revealed postoperative worsening EMA depression in 86 (7%), no significant change in 854 (73%), and significant improvement in 228 (20%). There were 40 patients (3.4% of 1166) with both worsening anxiety and worsening depression. Univariate predictors of postoperative worsening EMA depression included younger age, self-reported history of depression, preoperative pain, chronic pain, and higher preoperative PROMIS depression (Table 3). In the adjusted multivariable logistic regression model, significant predictors of postoperative worsening EMA depression symptoms included mild preoperative PROMIS depression symptoms (aOR compared with no baseline depression 2.26, 95% CI 1.24–4.14, P=0.008) and moderate to severe preoperative PROMIS depression symptoms (aOR 3.79, 95% CI 2.10–6.81, P<0.001). Self-reported history of depression was not a significant predictor of postoperative worsening EMA depression after adjusting for preoperative PROMIS depression. Results were similar in sensitivity analyses reclassifying patients with persistent severe anxiety as experiencing the outcome (Supplementary Table S3) and reclassifying 5% of patients with the smallest increases in postoperative anxiety as not experiencing the outcome (Supplementary Table S5).

Table 3.

Predictors of postoperative worsening depression EMA symptoms. CI, credible interval; EMA, Ecological Momentary Assessment; PROMIS, Patient-Reported Outcomes Measurement Information System. ∗Adjusted for all other variables shown in the table. The multivariable logistic regression model had a C-statistic of 0.74 and demonstrated a good fit using the Hosmer–Lemeshow test: χ²(8)=9.07, P=0.34. ^†t-Scores converted to categorical variables: ‘none to slight’ (t-score <55.0), ‘mild’ (t-score 55.0–59.9), ‘moderate’ (t-score 60.0–69.9), or ‘severe’ (t-score ≥70.0).

Variable	With worsening depression (n=86)	Without worsening depression (n=1082)	Unadjusted odds ratio (95% CI)	Unadjusted P-value	Adjusted odds ratio∗ (95% CI)	Adjusted P-value
Sex, n (%)
Female	61 (71)	771 (71)	(reference)		(reference)
Male	25 (29)	311 (29)	1.01 (0.62–1.64)	0.95	1.32 (0.77–2.24)	0.31
Age (yr), median (IQR)	49 (39–58)	54 (42–64)	0.98 (0.96–0.99)	0.006	0.99 (0.97–1.00)	0.13
Type of surgery, n (%)
Abdominal	42 (49)	544 (50)	(reference)		(reference)
Head, neck, and spine	22 (26)	171 (16)	1.67(0.97–2.87)	0.07	1.31 (0.73–2.36)	0.37
Orthopaedic	12 (14)	196 (18)	0.79 (0.41–1.54)	0.49	0.80 (0.39–1.64)	0.54
Other	10 (12)	171 (16)	0.76 (0.37–1.54)	0.44	0.63 (0.30–1.33)	0.23
Self-reported history of depression, n (%)	48 (56)	379 (35)	2.38 (1.52–3.71)	<0.001	1.44 (0.87–2.40)	0.16
Chronic pain, n (%)	33 (39)	300 (28)	1.63 (1.03–2.57)	0.03	0.99 (0.59–1.64)	0.96
Preoperative current pain, n (%)	77 (90)	798 (74)	2.98 (1.47–6.02)	0.002	2.07 (0.97–4.43)	0.06
Preoperative PROMIS depression†, n (%)
None to slight	37 (43)	812 (75)	(reference)		(reference)
Mild	20 (23)	150 (14)	2.92 (1.65–5.18)	<0.001	2.40 (1.33–4.35)	0.004
Moderate to severe	29 (34)	114 (11)	5.58 (3.30–9.43)	<0.001	3.97 (2.22–7.11)	<0.001

Discussion

In this single-centre, prospective, observational cohort study, we found that 5% of patients experienced postoperative worsening EMA anxiety symptoms during the first 30 days after surgery as measured using thrice-daily EMA, whereas 7% of patients experienced postoperative worsening EMA depression symptoms. Significant preoperative predictors of postoperative worsening EMA anxiety included preoperative pain and preoperative PROMIS anxiety symptoms. Significant preoperative predictors of postoperative worsening EMA depression included preoperative PROMIS depression symptoms. Notably, self-reported history of anxiety was not an independent predictor of postoperative worsening EMA anxiety, and self-reported history of depression was not an independent predictor of postoperative worsening EMA depression.

These findings emphasise the importance of assessing preoperative mental health symptoms, which were more strongly predictive of postoperative worsening anxiety and depression compared with self-reported history of these conditions. More severe preoperative PROMIS depression symptoms were more strongly predictive of postoperative worsening, whereas no similar severity-based pattern was seen with preoperative PROMIS anxiety symptoms. This finding can be attributed to various factors. Individuals with elevated preoperative levels of anxiety might continue to experience consistently high levels of anxiety after the postoperative period without significant worsening or improvement. Alternatively, preoperative anxiety symptoms might partially reflect situational anxiety related to the impending surgery, which might abate once the surgery has been completed. It is difficult to compare the lack of dose–response to previous studies, as, to our knowledge, no previous studies have examined preoperative anxiety and depression severity (mild or moderate to severe) as independent predictors of postoperative anxiety and depression. However, previous studies have similarly reported that preoperative anxiety and depression are risk factors for postoperative anxiety and depression.^21,22

The other identified predictors of postoperative worsening EMA anxiety and depression are generally consistent with past studies. Preoperative pain has been reported to increase the risk for postoperative anxiety.^22,23 Our findings are similar to previous studies that indicated no significant effect of age on postoperative anxiety.^8,24 However, younger age has previously been associated with increased postsurgical pain,²⁵ which might be pertinent given the observed association between preoperative pain and postoperative anxiety. Previous nonsurgical cohorts have also reported higher rates of depression in younger adults than in older adults.²⁶ Interestingly, the absence of associations between postoperative worsening anxiety or depression and chronic pain challenges conventional assumptions about the relationship between chronic illness and psychological distress.^11,23,24 This finding suggests that the psychological response to surgery might be influenced more significantly by individual differences in preoperative mental status rather than the presence of chronic medical conditions.

Although a significant percentage of individuals had worsening symptoms, the general trend of the cohort was for improvement in both EMA depression and anxiety symptoms. Use of EMA provided a high-resolution understanding of patients' emotional wellbeing, capturing momentary experiences, behaviours, and environmental factors.²⁷ This trend was also reported in a previous study examining anxiety and depression symptoms longitudinally after surgery.²³ Longitudinal studies of postsurgery recovery generally rely on data collected via retrospective recall (e.g. anxiety and depression symptoms over the past week). In this study, we used EMA to assess symptoms at the moment. This approach limits recall bias, which is known to impact reports of mental health symptoms.²⁸

Increases in anxiety, and even mild elevations in depression, might not always be detrimental. Major operations and transitioning from a hospital setting to everyday life can naturally provoke mild anxiety, signifying an adjustment to normalcy. In these situations, a mild increase in anxiety might even be an adaptive response. However, one would hope that such increases are short-lived, and, in the case of our study, such transient increases would not result in a positive slope if anxiety or depression reduced within the 30-day window. Because our EMA questions focused solely on symptoms and not on the effect of symptoms on daily activities, we cannot determine whether the symptoms we observed were adaptive or maladaptive.

A strength of our study lies in its large sample size and the diverse range of surgeries included, making it one of the few studies to assess postoperative depression and anxiety across various surgical procedures. In addition, the high temporal resolution of EMA data collection allowed our findings to be relatively resilient to transient changes in patient mental state compared with other scales. Our study maintained a high response rate and follow-up time during the postoperative EMA observation period, allowing for a thorough understanding of patients' mental states during the observation period. Lastly, our findings were stable during sensitivity analysis, indicating that they were not heavily influenced by our operational definition of the primary outcome.

Our study also has limitations. Firstly, EMA relies on self-reported data, which, similar to most measurements of patient wellbeing, can be influenced by various factors such as confirmation bias and survey fatigue. Secondly, anxiety and depression symptoms fluctuate, with EMA responses being dependent on the time of day at which they are collected. However, participants in our study completed the EMA multiple times per day, allowing general trends to be observed despite these fluctuations. Thirdly, our study cohort consisted of English-speaking patients with smartphone access within a single hospital system, perhaps limiting our findings to other clinical settings and patient populations. Fourthly, EMA is not an established scale with a validated minimal clinically important difference. However, our EMA drew its questions from the PROMIS questionnaire, which has been validated to capture a meaningful difference in patient symptoms accurately.29, 30, 31 Furthermore, our results remained stable in a sensitivity analysis reclassifying the bottom 5% of patient EMA slopes as representing no worsening of symptoms. Fifthly, our study used patient-reported histories of anxiety and depression, which might not have been accurately provided. Sixthly, our study only evaluated the role of preoperative predictors on worsening postoperative EMA depression and anxiety. However, intraoperative and postoperative factors might have also played a role in patients' postoperative mental wellbeing. Seventhly, some patients might have declined to participate in this study owing to hesitancy to disclose mental health conditions, potentially excluding some with a significant history of anxiety and depression. However, this concern is mitigated by the fact that the parent P5 study focused primarily on pain rather than mental health.

In conclusion, we demonstrated a 5% incidence of postoperative worsening anxiety symptoms and 7% incidence of postoperative worsening depression symptoms as measured by EMA during 30 days after major surgery. Severity of preoperative anxiety and depression symptoms were the most important predictors of postoperative worsening. Our findings demonstrate the importance of screening patients for mental health symptoms before and after surgery, which can facilitate development and implementation of tailored interventions to mitigate postoperative anxiety and depression.

Authors’ contributions

Study conception: SH, BAF

Study design: EA, KJH, TLR, SH, BAF

Statistical analysis: EA, BAF

Acquisition of data: MF

Collection of data: SH

Interpretation of data: EA, KJH, MF, TLR, CJ, SH, BAF

Initial drafting of the manuscript: EA

Critical revision of the manuscript: KJH, MF, TLR, CJ, SH, BAF

Declaration of interest

MF discloses consulting fees from Blueprint Health. The other authors declare no conflict of interest.

Funding

The grant was awarded to Dr. Haroutounian US Department of Defense (W81XWH-21-1-0736 to SH).

Handling Editor: Hugh C Hemmings Jr

Appendix A. Supplementary data

The following is the Supplementary data to this article: