Core temperature during surgery

logistic regression
survival analysis
Is low core temperature during surgery associated with poor surgical outcomes or death? Use logistic and survival analysis to study an observational dataset.
Author

Alex Reinhart

Published

May 17, 2023

Data files
Data year

2020

Motivation

During surgery that uses general anesthesia, patients often can have their body temperatures drop significantly. Anesthetics disrupt the body’s automatic mechanisms for maintaining body temperature, and they also tend to reduce the metabolism, meaning the body uses less energy and produces less heat. This can produce what is called “surgical hypothermia” or “inadvertent perioperative hypothermia” (IPH), when the body temperature drops below 36°C (96.8°F) during surgery. IPH can reduce the effectiveness of drugs, increase surgical bleeding, and delay recovery from surgery.

Researchers wanted to know whether IPH was also associated with surgical site infections (SSIs), which are also a serious problem during surgery. A surgical site infection can cause a patient to stay in the hospital longer or even lead to death. Some research suggested that IPH could increase the risk of SSIs, and so a new study explored whether there was an association between IPH and SSIs in patients who had colorectal surgery.

Data

The dataset includes 7,908 patients who had colorectal surgery at the Cleveland Clinic between 2005 and 2014. Patients were included in the study if their surgery took more than one hour and required general anesthesia, and if their esophogeal core temperature was monitored and sufficient baseline and outcome data was collected.

Data preview

core-temperature.csv

Variable descriptions

The variables include a number of baseline features about the patient and their health before surgery. They also categorize different types of infections after surgery, so specific infections can be analyzed individually, or we can consider all serious or superficial infections together.

Variable Description
YEAR Year of surgery
Age Patient’s age at the time of the surgery
FEMALE Patient’s sex at birth (1 = female, 0 = male)
BMI Patient’s body mass index (BMI) at time of surgery
CharlsonScore Patient’s Charlson Comorbidity Index score (higher numbers indicate more comorbidities that could lead to death)
SurgeryType This was meant to record the type of surgery (such as colostomy, ileostomy, or other procedures), but is missing for all patients
CHF Does the patient have a history of congestive heart failure? (1 = yes, 0 = no)
VALVE Does the patient have a history of peripheral vascular disease? (1 = yes, 0 = no)
DM Does the patient have a history of diabetes without chronic complications? (1 = yes, 0 = no)
RENLFAIL Does the patient have a history of renal failure? (1 = yes, 0 = no)
LIVER Does the patient have a history of liver disease? (1 = yes, 0 = no)
METS Does the patient have cancer that has metastasized? (1 = yes, 0 = no)
TUMOR Does the patient have a solid tumor that has not metastasized? (1 = yes, 0 = no)
COAG Does the patient have coagulopathy (a blood clotting disorder)? (1 = yes, 0 = no)
OBESE Is the patient obese? (Definition not specified; 1 = yes, 0 = no)
WGHTLOSS Has the patient lost weight? (Time period and amount not specified; 1 = yes, 0 = no)
LYTES Does the patient have fluid or electrolyte disorders? (1 = yes, 0 = no)
BLDLOSS Does the patient have chronic blood loss anemia? (1 = yes, 0 = no)
ANEMDEF Does the patient have a deficiency anemia? (1 = yes, 0 = no)
DRUG Does the patient abuse drugs? (1 = yes, 0 = no)
SteroidHx Did the patient use steroid drugs before surgery? (1 = yes, 0 = no)
ImmunosuppressantHx Did the patient use immunosuppressive drugs before surgery? (1 = yes, 0 = no)
SurgDuration Duration of the surgery, in minutes
Open Type of surgery. 1 = open surgery, 0 = laparoscopic (“minimally invasive”) surgery.
AbsessIntraAb Did the patient have an intra-abdominal abscess (caused by infection) after the surgery? (1 = yes, 0 = no)
AbsessPelvic Did the patient have a pelvic abscess (caused by infection) after the surgery? (1 = yes, 0 = no)
Cdiff Did the patient have a Clostridium difficile infection after the surgery? (1 = yes, 0 = no)
FascialDehiscence Did the patient suffer fascial dehiscence after surgery? This occurs when the fascia, which surround the organs in the abdomen and holds them in place, fails to heal after surgery and splits open. (1 = yes, 0 = no)
DelayedHealing Was the patient’s healing after surgery delayed? (1 = yes, 0 = no)
Infection Did the patient have a post-surgical infection that doesn’t fit into the other categories? (1 = yes, 0 = no)
Sinus Did the patient develop an abdominal sinus after surgery? (1 = yes, 0 = no)
SSIDeep(fascia) Did the patient develop a surgical site infection deep in the fascia surrounding the organs? (1 = yes, 0 = no)
SSIOrganSpace Did the patient develop a surgical site infection in the space surrounding the organs? (1 = yes, 0 = no)
SSISuperficial(skin) Did the patient develop a superficial surgical site infection in the skin around the incision? (1 = yes, 0 = no)
WoundInfection Did the patient develop an infection in the wound created in surgery? (1 = yes, 0 = no)
TWATemp Time weighted average of the patient’s core temperature during surgery, °C
LastReadingTemp Last recorded core temperature prior to surgery, °C
EndCaseTemp Patient’s core temperature at the end of surgery, °C
AnyInfection Did the patient develop any kind of infection, serious or superficial, within 30 days of surgery? (1 = yes, 0 = no)
SeriousInfection Did the patient develop a serious infection within 30 days of surgery? This includes the infections in the SSIDeep(fascia), SSIOrganSpace, AbsessIntraAb, AbsessPelvic, Cdiff, Pneumonia, Pneumonia(aspiration), and Sepsis variables. (1 = yes, 0 = no)
SuperficialInfection Did the patient develop a superficial infection within 30 days of surgery? This includes the infections in the SSISuperficial(skin), WoundInfection, and FascialDehiscence variables, as well as perineal wound problems. (1 = yes, 0 = no)
DurationHosp Time the patient stayed in the hospital after surgery, in days
LOS Time the patient stayed in the hospital, including time before surgery, in days
DEAD Did the patient die in the hospital? (1 = yes, 0 = no)

Questions

Introductory

  1. What kind of study is this: experimental or observational? What kinds of conclusions could we draw from the results, and what limitations would there be on our ability to draw causal conclusions?
  2. Plot the distribution of core temperatures during surgery (TWATemp). Comment on the shape of the distribution and find the median.
  3. Split the patients into two groups: those with temperatures above the median, and those with temperatures below the median. Calculate the proportion of patients in each group who had serious infections (SeriousInfection), and test whether there is a statistically significant difference.
  4. Compare these groups in other ways. Do they have different baseline health characteristics, such as age, BMI, and problems like liver or heart disease?

Advanced regression

  1. Fit a logistic regression model to predict serious post-operative infections (SeriousInfection) as a function of core temperature during surgery (TWATemp). Interpret the results of this fit.
  2. Extend your model to control for the covariates used in the original study: “age, sex, body-mass index, smoking status, Charlson score, duration of surgery, open (vs. laparoscopic), preoperative steroid usage, immunosuppressive drug usage,” and presence of heart failure, vascular disease, diabetes, renal failure, metastatic cancer, solid tumor, coagulopathy, weight loss, fluid and electrolyte disorders, blood loss anemia, deficiency anemia, and drug use. Interpret the results of this model, and compare it to the previous fit. What can you conclude about core temperature during surgery?
  3. Repeat this logistic regression, but to predict superficial infection instead of serious infection (SuperficialInfection). Comment on your results and compare them to the results for serious infection.

Survival analysis

  1. Split the patients into two groups: those with surgical core temperatures above the median, and those with temperatures below the median (using TWATemp). Plot estimated survival curves for the length of time the patient spent in the hospital after surgery (DurationHosp). Patients who died in the hospital are censored; the original study analyzed them as follows:

    The patients who died before hospital discharge were considered as never having the event and assigned a censoring time using the observed longest duration among those discharged alive.

    Interpret your survival curves in the context of the problem. Which group appears to spend longer in the hospital?

  2. Use a Cox proportional hazard model to study the association between core temperature during surgery (TWATemp) and time spent in the hospital after surgery (DurationHosp). Control for the confounding variables listed above (for the logistic regression analysis). Interpret your results.

References

Data first adapted for classroom use by: Homoki, J. and Nowacki AS (2023). “Core Temperature Dataset”, TSHS Resources Portal. Available at https://www.causeweb.org/tshs/core-temperature/. Released under a CC BY-NC-SA license.

Original study: Walters, Michael J., et al. (2020). “Intraoperative Core Temperature and Infectious Complications after Colorectal Surgery: A Registry Analysis.” Journal of Clinical Anesthesia, vol. 63. https://doi.org/10.1016/j.jclinane.2020.109758

Further information on surgical hypothermia, its causes, and its effects: Riley C, Andrzejowski J. (2018). “Inadvertent perioperative hypothermia.” BJA Education 18(8), 227-233. https://doi.org/10.1016%2Fj.bjae.2018.05.003. PubMed Central.