Methods
Data Sources
To examine the relationship between preoperative psychiatric diagnoses and treatment on postoperative outcomes, we linked the Patient Care Services databases (Corporate Data Warehouse [CDW]) of the Veteran Health Administration (VHA) to data from the U.S. Department of Veterans Affairs (VA) Surgical Quality Improvement Program (VASQIP). CDW data are derived from the VHA’s national electronic medical record and capture health care encounters (
10). VASQIP is a mandatory national program designed to measure and provide feedback on the quality of surgical care at VHA hospitals (
11). VASQIP includes both preoperative (e.g., patient demographic characteristics, comorbid conditions, and laboratory values) and postoperative outcome data. Previous work has demonstrated the accuracy and reliability of the data (
12).
This study was approved by the Baylor College of Medicine Institutional Review Board and the Michael E. DeBakey VA Medical Center Research and Development Committee. Data analyses took place between 2017 and 2019.
Sample Population and Database Linkage
In this study, we identified all patients in VASQIP who underwent colon or rectal resection (Current Procedural Terminology [CPT] codes 44140–44160, 44204–44212, 45110–45149, 45151–45172, 45395, and 45397) from 2000 to 2014 in conjunction with a primary diagnosis code for colon or rectal cancer (ICD-9 codes 153.0–153.4, 153.6–153.9, 154.0–154.3, 154.8, 230.3, and 230.4). This screening and selection resulted in a retrospective cohort of 58,961 patients. VASQIP records were then linked to each patient’s CDW data to identify preoperative psychiatric diagnoses (according to ICD-9 codes) and receipt of preoperative mental health treatment (CPT codes; see online supplement for both codes).
Variables
The main variables of interest were presence of a psychiatric diagnosis, documented by any VA provider within 30 days before surgery, and receipt of mental health treatment within 60 days before surgery. Outcome variables included occurrence of a postoperative complication, defined per VASQIP records as perioperative mortality, cardiac complications, postoperative pneumonia, intubation for >48-hour postoperatively unplanned reintubation, venous thromboembolic events, renal dysfunction, or surgical site infections; increased postoperative hospital LOS; and readmission within 90 days of surgery.
Statistical Analyses
We performed multivariable analyses to examine the association between preoperative psychiatric diagnoses and treatment on postoperative outcomes; we used generalized estimating equations to account for clustering of patients at the surgical facility. We modeled complications and readmissions by using a binary distribution and modeled LOS with a Poisson distribution.
Subgroup analyses were conducted to examine the association of each psychiatric diagnostic category (i.e., depressive, anxiety, bipolar, posttraumatic stress, substance use, psychotic, cognitive, and personality disorders), compared with having no preoperative psychiatric diagnosis, with postoperative outcomes. Subgroup analyses examining the association of preoperative mental health treatment on postoperative outcomes for each diagnostic category were not conducted because of insufficient statistical power. However, we computed the percentages of complications and hospital readmissions, as well as median LOS by treatment type, for each diagnostic category.
Model covariates included demographic factors (age, sex, race, hospital region, rurality, and year of surgery) and preoperative clinical factors, including the Deyo modification of the Charlson comorbidity index (
13); current smoking status; functional status (independent, partially or totally dependent); American Society of Anesthesiology classification; systemic sepsis or ventilator dependency 48 hours before surgery; steroid use for a chronic condition, ascites, or chemotherapy within 30 days before surgery; radiotherapy within 90 days of surgery; 10% loss of body weight in the 6 months preceding surgery; preoperative serum creatine; postgraduate year of surgeon; emergency surgery status; surgical approach (laparoscopic, open, or both); site (colon, rectum, or both); presence of an ostomy; and postoperative wound class (clean, contaminated, or infected). We conducted all analyses with SAS, version 9.4, using the PROC GENMOD procedure.
Discussion
Numerous quality-improvement efforts focused on improving surgical care outcomes in the United States are ongoing (
2,
3). Although many factors have been identified as being associated with worse postoperative outcomes, few can be modified before surgery. Mental health problems among surgical patients remain understudied, but data suggest that they may represent an underappreciated risk factor for poor postoperative outcomes and that they may be responsive to preoperative treatment, which may, in turn, improve surgical outcomes. In this regard, our work supports two main conclusions. First, preoperative psychiatric diagnoses are a risk factor for worse surgical outcomes, and this risk may depend on the particular psychiatric diagnosis. Specifically, substance use disorders were associated with increased risk for postoperative complications; depression and psychotic disorders were associated with longer LOS; and depressive, posttraumatic stress, and substance use disorders were associated with an increased risk for 90-day readmission. The association of substance use disorders, which were largely accounted for by tobacco use disorders, with postoperative complications and increased readmission rates was not surprising given tobacco’s detrimental effects on immune, cardiac, and respiratory functions important for wound healing (
14).
Similarly, depression has long been associated with increased LOS among medical-surgical inpatients (
7,
15,
16) and with postoperative readmissions (
16–
18). The association of depression with LOS and readmission is likely due to the systemic effects of depression on neurobiological processes associated with wound healing (e.g., increased inflammation [
19,
20] and platelet reactivity [
21] and decreased heart rate variability [
22]), and because depression is associated with reduced social support (
23), which can contribute to poorer postoperative recovery (
24). Unlike previous studies focused on elderly surgical patients, which found that LOS is unaffected by psychiatric disorders except for cognitive disorders (
25,
26), we did not observe a significant association between cognitive disorders and postoperative outcomes. This finding was likely due to the relatively small portion of the sample with cognitive disorders (0.4%, N=239). Similarly, the lack of statistically significant findings for other preoperative diagnoses (i.e., personality, bipolar, and anxiety disorders) was likely due to the relatively smaller sample sizes for these diagnostic categories. Future research with larger sample sizes may clarify the association of different diagnoses with postoperative outcomes.
Second, the findings of this study suggest that the risk associated with a psychiatric diagnosis may partially depend on preoperative mental health treatment. Specifically, patients who received psychotherapy only during the month before surgery experienced postoperative outcomes similar to those without a psychiatric diagnosis, whereas patients who received no mental health treatment or only psychiatric medication had an increased risk for postoperative complication and readmission compared with those without a psychiatric diagnosis. However, because of the small number of patients with a psychiatric disorder who received preoperative psychotherapy in this cohort (N=308), we cannot draw a firm conclusion from these data about the association of preoperative psychotherapy with postoperative outcomes.
Additionally, patients receiving no mental health treatment or psychiatric medication alone may systematically differ from those engaging in psychotherapy (e.g., those receiving psychiatric medication may have more severe psychiatric symptoms). Indeed, psychiatric symptom severity predicts poorer postoperative outcomes (
27), whereas receipt of psychiatric medication is associated with improved postoperative outcomes (
28,
29). Future research with a larger sample of participants receiving preoperative mental health care may help clarify these associations. Interestingly, contrary to previous research (
25,
26,
30), LOS did not substantially differ between patients with and without psychiatric diagnoses, regardless of preoperative mental health treatment.
Although the findings related to the impact of preoperative mental health care on postoperative outcomes are far from conclusive, observations from previous research suggest that preoperative psychological and psychiatric care are both associated with improved postoperative outcomes. For example, four sessions of cognitive-behavioral therapy before coronary surgery led to a 1-day shorter postoperative LOS compared with usual care for patients with depression (
31). Similarly, two sessions of stress management before breast and prostate cancer surgery improved postoperative health-related quality of life (
32,
33) and immune functioning critical to wound healing (
34,
35), especially for patients with distress (
36). A “prehabilitation” program before colorectal surgery, including exercise, nutritional counseling, and psychologist-led relaxation, led to better postoperative walking capacity compared with standard care (
37), and effects were particularly strong for patients with anxiety (
38). Preoperative counseling and nicotine replacement for 4–8 weeks were associated with fewer complications and shorter LOS for tobacco users undergoing surgery (
39,
40). Additionally, proactive psychiatric consultation has been associated with a 1-day shorter LOS among general hospital patients (
28,
29). Preoperative mental health care also fits with national calls for integrating psychosocial care into oncology care (
41,
42). In summary, the present study’s findings may be considered hypothesis generating, warranting future studies examining the effects of preoperative mental health care (both psychotherapy and psychiatric care), delivered in randomized controlled trials and real-world contexts, on postoperative outcomes.
Although 15% of patients in our cohort had a documented psychiatric diagnosis at the time of surgery, this percentage likely represents an underestimate of the prevalence of mental health problems among surgical patients. For example, in primary care settings, which routinely screen for psychiatric disorders, >60% of patients who met criteria for a psychiatric disorder did not have such disorder documented in their medical record (
8,
9). Most patients (76%) with advanced cancer who met criteria for a psychiatric disorder reported that they had not discussed mental health with a medical provider (
43), and cancer providers often fail to recognize patients’ mental health needs (
44,
45).
Careful screening for mental health concerns during preoperative appointments may help identify patients at risk for postoperative complications and readmissions; moreover, it would fulfill the American College of Surgeons Commission on Cancer’s accreditation requirement to screen for distress at critical junctures in patients’ oncology treatment (
41,
46). Increased accuracy in detecting psychiatric diagnoses would likely reveal an even greater impact of mental health on postoperative outcomes. For example, whereas this retrospective study found that psychiatric diagnoses were associated with ORs of 1.09 and 1.11 for complications and readmissions, respectively one prospective study found that preoperative symptoms of depression, anxiety, or substance use disorders were associated with a much greater OR of 3.75 for 30-day readmissions compared with patients without psychiatric symptoms (
47).
Effective mental health screening and appropriate follow-up occur more regularly when a mental health provider is embedded, or colocated, in the surgical oncology clinic. Colocation can remove barriers to care, including scheduling, transportation, stigma, and miscommunications. For example, the presence of a proactive psychiatric consultation service doubled the rate of consultations, increased the psychiatric diagnoses detected, and significantly reduced LOS among medical inpatients (
28,
29). Although preoperative mental health care is costly, and including a mental health specialist on the treatment team can be challenging, patients (
43) and providers have reported the desire for greater integration (
48); furthermore, the return on this investment may be substantial in light of the high quality of life and financial cost of in-hospital stays, complications, and readmissions after surgery.
This study has several important limitations. First, the findings related to the association between mental health treatment type and postoperative outcomes should be interpreted with caution because so few patients received psychotherapy with (N=136) or without (N=308) psychiatric medication. Additionally, we found that psychiatric diagnoses were associated with small increases in the risk for complication (OR=1.9), LOS (rate ratio=1.04), and readmission (OR=1.11), independently of covariates. However, even small increases in these outcomes can have serious implications for quality of life and cost.
Moreover, our administrative databases did not capture data on postoperative pain and psychiatric medications prescribed for reasons unrelated to mental health (e.g., tricyclic antidepressant for chronic pain) and did not include postoperative outpatient visits as a potential outcome variable. In addition, these databases did not allow consideration of the severity of psychiatric diagnoses. It therefore is possible that receipt of psychiatric medication may be a proxy for psychiatric disorder severity, which may explain the relatively worse outcomes for patients receiving psychiatric medications. The available data also did not identify the type of substances used by patients with a substance use disorder, which may be particularly informative in this patient population, given that half of the patients in this cohort with a psychiatric diagnosis also had a diagnosis of substance use disorder. Furthermore, substance use disorders included tobacco use disorders, which are known to directly affect postoperative recovery.
Finally, this study focused on a narrow 1-month window before surgery, precluding consideration of previous psychiatric diagnoses or treatments. This study also included data solely from VHA databases. Patients may have accessed mental health or postoperative care outside the VHA. However
, a previous study examining both VHA and Medicaid-Medicare data found that most (>90%) VHA patients received postoperative care at a VHA facility (
49). Additionally, it is important to examine these associations in a non-VA setting to assess generalizability.