As part of our institute’s “critical pathways”, we organized a multidisciplinary team of experts, including surgeons, dietitians, endocrinologists, anesthesiologists, physiatrists, psychiatrists, and psychologists, all having substantial experience in supportive care of cancer patients. Each specialization provided personalized care e.g., nutritional support, diabetes control, pain management, preoperative and postoperative rehabilitation, and psychiatric intervention to patients before and after pancreatectomy according to its own management protocols (S1 Fig.). Surgeons were responsible for the integrated management of the patients and made referrals to relevant specialists upon problem detection.

For psychological distress management, we developed a distress management protocol for patients undergoing pancreatoduodenectomy based on the Korean distress management guidelines [11]. This protocol included a routine screening and systematic psychiatric referral system for psychological symptoms based on the two-tiered and triage model. The patients’ psychological symptoms were categorized into three domains (insomnia, anxiety, and depression), based on commonly experienced psychiatric symptoms in cancer patients [12]. Patients were routinely screened for symptoms at the time of admission (T1) and on the tenth postoperative day (T2). For those who had psychological symptoms at T2, an in-depth evaluation of their symptoms was performed on the same day. The screening and in-depth questionnaire were administered by a registered nurse, who assisted the patients in reading, understanding, and completing the questionnaire. Then, a clinical psychologist interpreted the test score, decided the pathway for care according to the algorithm, and recommended aftercare to the surgeon. The clinical psychologist recommended emotional support and the provision of educational medical information by the primary medical staff to patients with normal to mild distress levels. Referral for psychiatric intervention was provided to patients with mild-to-severe levels of distress. For patients who did not complete the in-depth evaluation, whether to refer to mental health services was decided based on the results of the T2 screening. Surgeons referred those with clinically significant symptoms to psychiatric consultation for receiving relevant management. The psychiatrists and/or psychologists visited the patients for examination and consultation. Patients who were referred to the mental health clinic received pharmacological and/or non-pharmacological intervention, such as psychotherapy and psychoeducation. The median hospital stay of the study population was 25 days (range, 12 to 123 days); hence, all the eligible patients were in admission when T2 screening was performed and the in-depth questionnaire was administered. The surgeons were responsible for supervising the registered nurse and making referrals to the psychiatrists and psychologists. A screening and consultation request system was implemented on the electronic medical records (EMR) system to enable communication among the experts. Screening results at T1 mainly served as the baseline, but occasionally, the patient was referred to a psychiatrist when it was determined that psychiatric treatment was required.

In this retrospective study, de-identified archival data was analyzed, while the serial data of patients who underwent a pancreatoduodenectomy for a periampullary neoplasm (from September 2010 to December 2018) were retrospectively reviewed. Patients whose operations were canceled, or who did not otherwise undergo planned pancreatoduodenectomy, were excluded from analysis. Among the 440 patients admitted for the procedure during the study period, 364 patients who completed screening of psychological symptoms at T1 and underwent pancreatoduodenectomy were included in the analysis.

Screening was performed at T1 and T2 by measuring psychological symptoms with the Korean version of the self-reported National Cancer Center Psychological Symptom Inventory (NCC-PSI) [12]. The NCC-PSI consists of six items on an eleven-point visual analogue scale that measures the severity and impact of insomnia, anxiety, or depression over the past week. One item enquires about the “need for help from mental health experts” for each symptom [12]. The cutoff score in the NCC-PSI for clinically significant insomnia, anxiety, and depression was five, four, and four points, respectively [12].

In-depth questionnaires were administered to patients who had NCC-PSI scores higher than the cutoff score at T2 to triage patients who needed priority psychiatric intervention. The severity of insomnia was measured using the Korean version of the Insomnia Severity Index (ISI) [13,14]. Each of the seven items was rated on a five-point Likert scale (0, not at all; 4, very severe), and the range of total scores for insomnia was interpreted as not being clinically significant (0–7 points), mild (8–14 points), moderate (15–21 points), or severe (22–28 points) [14]. Meanwhile, the severity of state anxiety was measured using the Korean version of the self-reported State-Trait Anxiety Inventory (STAI-X) [15,16]. Each of the 20 items was rated on a four-point Likert scale (1, almost never; 4, almost always), and the results for the patient’s anxious state were classified as insignificant (20–51 points), mild (52–56 points), moderate (57–61 points), or severe (62–80 points) [16]. Finally, the level of depression over the past week was measured using the Korean version of the self-reported Beck Depression Inventory (BDI) [17,18]. Each of the 21 items was rated via four possible responses (0–3), and the severity of depression was categorized as insignificant (0–9 points), mild (10–15 points), moderate (16–23 points), or severe (24–63 points) [18]. By utilizing the cutoff score suggested by the scale, questionnaire results rated as mild-to-severe were considered clinically significant.

The demographic and clinical data from the EMR were retrospectively collected, including the following: age; sex; the status of marriage, employment, and education; smoking and alcohol intake; past psychiatric history; medication; family history of cancer; insight into one’s disease; cancer type; and the European Cooperative Oncology Group (ECOG) performance scale at the time of hospitalization. In addition, the following data were collected: patients’ nutritional status (via a patient-generated subjective global assessment [PG-SGA]) at T1, any complications that developed within ten days after the operation, and the highest pain score on the 10th postoperative day (T2) as well as the postoperative hospital stay.

Continuous variables were expressed as median and interquartile range (IQR) and compared using Mann-Whitney U tests. The serial NCC-PSI scores at T1 and T2 were compared using Wilcoxon signed-rank tests. The categorical variables were compared using the chi-square or Fisher exact test, and the correlation was tested via Spearman’s correlation test. All the statistical analyses, which were performed using STATA ver. 16.1 (StataCorp LLC, College Station, TX), were two-sided, and the statistical significance was defined as p < 0.05.

The adherence rate for routine screening psychological symptoms at the time of admission was 82.7% (364/440) for all the patients who underwent a pancreatoduodenectomy during the study period. Of the 364 patients, 91.8% (n=334) completed their screening at T2. Response rates for the in-depth questionnaires were 80.7% (88/109, ISI), 75.0% (84/112, STAI-X), and 68.5% (63/92, BDI), respectively. Although all the questionnaires were administered one-on-one, patients tended to refuse to answer it due to their poor general condition or to drop out when the questionnaire was lengthy (e.g., STAI-X or BDI). In addition, psychiatric referrals were occasionally omitted because the referral process was not electronically automated and had to be handled by the surgeons (S2 Fig.).

The median annual number of patients was 45 (IQR, 37 to 48), and their demographics are presented in Table 1. The patients with distal common bile duct cancer included a higher proportion of men (80.2% [n=65] vs. 63.6% [n=96, pancreatic head], 55.4% [n=46, ampulla of Vater], 61.2% [n=30, other types]; p=0.007), and the proportion of patients with an ECOG score that was higher than or equal to two was higher for those with pancreatic head cancer (11.3% [n=17] vs. 2.5% [n=2, distal common bile duct], 7.2% [n=6, ampulla of Vater], 2.0% [n=1, other types]; p=0.039). Other demographic factors were comparable across the primary sites of disease.

The prevalence of insomnia, anxiety, or depression at T1 was 22.0% (n=80), 29.1% (n=106), and 18.4% (n=67), respectively, and after the operation, these percentages increased by 10.6%, 4.5%, and 9.2% (Fig. 1). At T1 and T2, 37.9% (138/364 [T1]) and 46.4% (155/334 [T2]) of patients had psychological symptoms in at least one of the three domains. However, less than 15% and 20% of all patients at T1 and T2, respectively, expressed their “need for help from mental health experts” for their psychological symptoms through the items on the screening questionnaire (NCC-PSI). Among those with scores higher than the cutoff, less than 45% expressed the need for psychological assistance, and this frequency was even lower in patients with depression, compared to those with insomnia or anxiety. The presence of psychological symptoms at T1 and T2 showed significant correlation (r_s=0.173, p=0.002 [insomnia]; r_s=0.352, p < 0.001 [anxiety]; r_s=0.353, p < 0.001 [depression]). Each domain illustrated significant cross-correlation (insomnia and anxiety, p < 0.001; insomnia and depression, p < 0.001; anxiety and depression, p < 0.001), especially the patients with depression who had concurrent anxiety, 92.5% (n=62) at T1 and 84.8% (n=78) at T2.

For the screening test at T2, the patients who exceeded the cutoff score across the three domains were evaluated in-depth for the relevant domain. The ISI was completed for 80.7% (88/109) of eligible patients, and of these, 77.2% (n=68) had a mild-to-severe level of insomnia. As shown in Fig. 2A, 28.4% of patients who did not have clinically significant insomnia at T1 developed symptoms after the operation. Meanwhile, the STAI-X was completed for 75.0% (84/112) of eligible patients, and of these, 38.1% (n=32) had either a mild-to-severe level of state anxiety. Among the patients with clinically significant anxiety at T1, 59.0% (n=59) remained symptomatic after the operation (Fig. 2B). The proportion of patients who had anxiety at T1 was higher than those who did not (85.7% [n=12] vs. 14.3% [n=2], p=0.011) among 14 patients with moderate-to-severe anxiety. Finally, BDI was completed for 68.5% (63/92) of eligible patients, of whom 82.5% (n=52) had either a mild or higher level of depression. Sixty percent of the patients who experienced clinically significant depressive symptoms at T1 remained symptomatic after the operation (Fig. 2C).

Related to the clinical level of psychological symptoms at T1 and T2, we analyzed pre- and postoperative factors listed in the “Demographic and clinical variables” section. At T1, insomnia was related to having a poor nutritional status, no previous history of alcohol intake, and a past history of depression/insomnia. Additionally, a history of depression was related to having anxiety/depression at T1 (Table 2).

The postoperative hospital stay was significantly longer for patients who had clinically relevant anxiety (median, 21 days [IQR 16–29] vs. 18 days [IQR 14–23]; p=0.001) or depression (median, 21 days [IQR 16–29] vs. 18 days [IQR 15–24]; p=0.007) at T2. Except for elevated serum C-reactive protein level in patients with clinically relevant anxiety (median, 8.5 [IQR 2.9, 13.4] vs. 4.7 [IQR 1.3, 10.6]; p=0.028), postoperative complications [any complications, Clavien-Dindo classification grade ≥ III, postoperative pancreatic fistula grade ≥ B, delayed gastric emptying grade ≥ B, postoperative hemorrhage requiring transfusion, superficial surgical site infection only], serum hemoglobin, serum C-reactive protein, and numeric rating scale of pain did not have statistical significance in relation to clinically relevant insomnia, anxiety or depression at T2.

Meanwhile, patients who had two or more symptoms at T2 had a longer postoperative hospital stay, as compared to those with one or no symptoms (a median of 20.5 [IQR, 16.0 to 29.0] vs. 18.0 [IQR, 14.3 to 24.0], p=0.006). Although, according to the number of symptoms, notably, the following factors were not significantly different: age (a median of 67.0 [IQR, 58.0 to 72.0] vs. 67.0 [IQR, 59.0 to 72.0], p=0.911) and complication grade (Clavien-Dindo classification ≥ grade 3, 45.1% [n=23] vs. 54.9% [n=28], p=0.344).

A multivariate analysis of the effect of psychological symptoms and postoperative complications on postoperative hospital stay revealed that clinically relevant anxiety at T2 (odds ratio, 4.930; 95% confidence interval [CI], 1.300 to 8.559; p=0.008), postoperative pancreatic fistula grade ≥ B (odds ratio, 13.115; 95% CI, 9.443 to 16.796; p < 0.001), and delayed gastric emptying grade ≥ B (odds ratio, 12.314; 95% CI, 8.036 to 16.591; p < 0.001) were significant factors related with postoperative hospital stay.

After the psychological symptoms screening at T2, 31.1% (n=104) of all patients had a psychiatric consultation, while 52.9% (n=82) who had psychological symptoms at T2 were provided this service as well. Overall, Among the 92 patients who had a mild-to-severe level of in-depth questionnaire scores, 72.8% (n=67) underwent psychiatric consultation.

Among the 104 patients who had in-patient referrals, 82.7% (n=86) and 57.7% (n=60) were treated with psychotherapy or medication, respectively. The median number of psychotherapy sessions was one (IQR, 1 to 2), while medication was prescribed for a median of 13 days (IQR, 6 to 29). The top three frequent in-hospital medications prescribed were zolpidem (28 patients; IQR, 3 to 10 days), quetiapine (19 patients; IQR, 12 to 32 days), and lorazepam (18 patients; IQR, 7 to 20 days). Twenty-seven patients required follow-ups at the out-patient clinic due to persistent symptoms; 100% (n=27) and 59.3% (n=16) received psychotherapy and medication, respectively. After discharge, the median number of psychotherapy sessions that were performed at the out-patient clinic was one (IQR, 1 to 3), while medication was prescribed for a median of 28 days (IQR, 22 to 43).