Surgery, anesthesiology, and emergency and critical care medicine

The Biostatistics Clinic on Wednesdays is dedicated to biostatistics applications in surgery, anesthesiology, and emergency and critical care medicine.

Click here for 2023, 2022, 2021, 2020, 2019, 2018, 2017, 2016, 2015, 2014, and before.

Current Notes (2024)

2024 September 11

Janelle Buysse (Kristina Betters), Pediatric Critical Care

Our study is a descriptive study assessing utility of an early mobility protocol in the pediatric cardiac ICU.

2024 September 4

Nora McNulty (Jeff Heimiller), Emergency Medicine Simulation

Simulation, particularly employing Rapid Cycle Deliberate Practice, has proven invaluable in teaching residents the dynamics of a well-executed code. It immerses them in the high-pressure environment of a cardiac arrest, allowing practice of ACLS skills in various roles repeatedly and at their own pace. It is our hope that RCDP can be adapted for self-directed learning. By providing access to a manikin and pre-programmed cases, residents can conduct their own cardiac arrest simulations and receive performance metrics generated by the simulation software. We believe these metrics will enable residents to assess their own performance and identify performance gaps without simulation faculty present.

Our hypothesis is that residents engaged in self-directed simulation will progress in their learning as effectively as those receiving traditional coaching from faculty methods, which is the standard approach. During a Super Tuesday simulation session with the residents, residents will be randomly assigned to two groups" faculty-guided and self-directed RCDP training. We will employ multiple objective and subjective measures to evaluate the residents’ skills post-session. Our aim is to compare the effectiveness of faculty-guided versus non-faculty-guided training and determine if the non-faculty-guided approach results in similar skill development.

Our question for the biostatistics clinic is this, we will only have 12 groups of residents total. Which means we will be only randomizing 6 groups into the faculty-driven group and 6 groups into the non-faculty driven group. What can we do to increase the power of our study in our data collection?

2024 August 28

Whitney Barham (Matthew Alexander), Cardiology

Observational, retrospective cohort study: “In patients with metastatic melanoma who are treated with ICI therapy (nivolumab, ipilimumab, or pembrolizumab), is a specific type of ICI treatment associated with elevated systolic blood pressure at 6 month or 2 years of follow-up?” - We plan to complete a data pull from the RD of melanoma patients treated with these medications and obtain additional data related to hypertension during the follow-up period. Based on a previous, smaller study, our hypothesis is that treatment with dual therapy (Ipilumumab+Nivolumab) as compared to single agent (Nivolumab or Pembrolizumab alone) would be associated with increased systolic BP at the 2-year follow-up time point.

2024 August 14

Jason Samuels, General Surgery

This is a proposal for the Learning Healthcare Systems Pragmatic Trials RFA. We are proposing a pragmatic randomized trial comparing ambulatory (same-day discharge) bariatric surgery to current standard of care (≥ 24 hour admission). The study in current iteration is a non-inferiority trial with the primary outcome of ER visits within 7 days of surgery. Estimates for ER visitation rate in current literature is 4% in our hospital-admission (control) arm compared to 8% in our same day discharge (intervention) arm, with an allowable margin of difference of 10%. We are seeking guidance on randomization strategies (clustered by month or surgeon versus patient-level randomization) as well as ensuring appropriate sample size.

Meeting Notes:

• Primary outcome: compare ER visits in first 7 days between the two groups (one night in hospital; same day discharge)
• Goal is a pragmatic trial with a waiver of consent

Recommendations:

• Think about the margin of difference. Is 10% too high?
• For sample size, estimate how many subjects will be able to be enrolled in the study time frame and justify this. Can estimate power from this
• Cross over study instead of a cluster randomization would be the best approach here (to avoid issues with differences in surgeon and protocol deviations)
• Look into the SMART and SALT-ED trial designs

2024 May 29

Anna Pfalzer, Neurology

I am currently running a proteomics platform on ~700 plasma samples from patients diagnosed with 30+ neurodevelopmental disorders and unaffected age-matched controls.

I need assistance with the initial across- disorder analyses - to determine the variability and/or clustering of all samples independent of diagnosis.

Meeting Notes:

• Not everyone in the study has an unaffected match; this could be a limitation

Recommendations:

• Include descriptive statistics by group
  • Start with one biomarker. Compare between affected and unaffected
  • Then, compare between all disorders and also unaffected
    • Can use Kruskal Wallis test to compare
  • Will need to adjust for multiple comparisons
• Can do a PCA within each proteomic group
• To include covariates, can use logistic regression
  • Outcome: affected vs unaffected
  • Primary predictor: proteomic biomarker
  • Adjust for covariates
• To compare the association of each covariate between affected and unaffected, interaction terms can be used to test differential treatment effects
• Recommended to apply for a VICTR voucher
  • Might be too large for 1 voucher (can mention in application that it might need 2 vouchers)
  • Include a statistical analysis plan and sample size justification

2024 May 22

Hayden Byrd (Austin Kirschner), Radiation Oncology

MOBILE is an ongoing multi-institutional study treating patients with osteoarthritis with radiation. The data being collected includes basic patient demographics such as age and gender, which joint(s) are affected, their baseline pain, and their pain response at time intervals from treatment. We would like to perform an interim analysis on patients treated thus far.

Meeting Notes:

• Would like to analyze pain response (pain scales, quality of life survey, weaning from pain drugs) stratified by patient demographics and site treated
• Evaluate which patients go for a second treatment
• Currently have approximately 75 patients enrolled, with a final goal of approximately 150

Recommendations:

• Clearly define the scope of the analysis to ensure that all work can be completed within the 90 hour limit of a VICTR voucher
• Suggest checking data quality in what has been collected so far. Data checking and cleaning can be done within the scope of the voucher but will take away from analysis time
• Dandan has confirmed that this project is a good fit for a VICTR voucher

2024 April 17

Kristina Betters, Pediatric Critical Care

Our goal is to assess patient and system factors associated with readmission to the PICU within 24 hours of transfer to acute care. We would like to discuss statistical analyses options prior to starting data collection.

Meeting Notes:

• Are there risk factors for patients that get transferred back to the PICU within 24 hours after being transferred out?
• Data will be stored in REDCap (REDCap has been created, but no data collection yet)
• Around N = 90 transferred back within 24 hours in a sample time frame that was pulled

Recommendations:

• If possible, use all data. If not, matching can be used
  • If using matching, consider what to match on
• Clearly define the outcome (consider how to include death in the outcome)
• Focus in on a few main risk factors/predictors (i.e. PEW score, time of day, respiratory patients) to develop hypotheses
• Need to consider how to deal with multiple encounters for a patient
  • First look at how often this happens
• Recommended to apply for a VICTR voucher

2024 March 6

Eesha Singh (Jillian Berkman), Neurology

Moyamoya vasculopathy (MMV) is a disease of the intracranial arteries leading to narrowing and occlusion. This contributes to increased stroke risk, with most patients being diagnosed with MMV after they have had a stroke.
MMV can occur in the setting of genetic predisposition, Moyamoya disease (MMD), or vascular remodeling secondary to another pathologic process, Moyamoya Syndrome (MMS). In the United States, MMS often occurs
concurrently with uncontrolled stroke risk factors such as hypertension, diabetes, or smoking.

The mainstay of treatment remains bypass surgery, with lack of significant data regarding medical management of MMS in the US. While there are some anecdotal or case series-based data, there is little to no representative data
examining the management of MMS patients, especially in the context of socioeconomic factors.

Social determinants of health contribute to risk factors that can increase risk for MMS and lead to poor management and outcomes. Anecdotally, patients with uncontrolled vascular risk factors and subsequent atherosclerosis do not reliably receive appropriate monitoring or treatment. We aim to assess the treatment and interventions that adult patients with symptomatic MMS in the Stroke Belt are offered and receive.

The severe limitation that sparse data and lack of guidelines in this population imposes on clinical practice requires that more studies and reviews to identify the unique risk factors contributing to MMS. We hope to evaluate the impact of individual as well as population-level social determinants, outlined below, on MMS management, both medical and surgical.

Meeting Notes:

• Main research question: How do SDOH affect clinical outcomes and management in Moyamoya disease?

Recommendations:

• Clearly define outcomes (clinical and management)
  • Clarify the time frame for follow up
• Consider any potential confounders
• Include medical history (ex. history of diabetes, history of hypertension, medications, etc.)
• Can consider a time-to-event analysis or logistic regression (primary outcome = recurrent stroke)
• Recommended to apply for a VICTR voucher

Andrew James (Patrick Assi), Plastic Surgery

We are looking to have a statistician review our statistical strategy for an upcoming publication and provide insight into additional considerations we may need to take into account.

Meeting Notes:

• Hypothesis: Exposure to testosterone does not increase the incidence of pathology
  • 80% of cohort are using hormone therapy

Recommendations:

• Stratify Table 1 by testosterone (yes/no)
  • Include an overall column
  • Include demographics, medical history, pathology results
  • Chi-squared tests (include p-value)
    • This is fine for an unadjusted analysis (but does not take order of pathology severity into account)
• Can collapse some of the levels for pathology with small cell counts
• In Table 2, include subjects with benign pathology, so the total adds up to the full cohort
• Use Fisher\x{2019}s exact when cell counts are less than 5

2024 January 31

Kevin Johnson, Pediatric Surgery

Our project utilizes a large database doing a time-to-event analysis on neurologic complications in ECMO patients. Our question is whether there is a statistical analysis that we can do to compare the risk of a neurologic complication when splitting the analysis around a time point (currently have Kaplan Meier curves).

Meeting Notes:

• If there is an established collaboration plan with the biostatistics department, this might be the best place to go (Dandan to make this connection)
• Otherwise, can apply for a VICTR voucher

Recommendations:

• Look at cumulative incidence plots to help identify what timeframe to look at
• Eventually will want to run a proportional hazards model, to include all potential risk factors in one model
  • Large event rate \x{2013} should not have a problem fitting everything in the model

2024 January 24

Kevin Johnson, Pediatric Surgery - No Show

Our project utilizes a large database doing a time-to-event analysis on neurologic complications in ECMO patients. Our question is whether there is a statistical analysis that we can do to compare the risk of a neurologic complication when splitting the analysis around a time point (currently have Kaplan Meier curves).

2024 January 17

Kristina Betters, Pediatrics Critical Care

We are implementing a diary for caregivers and patients to use in the PICU and would like to assess whether the diary is effective in helping patients and caregivers cope and reduce anxiety, stress and trauma related to hospitalization.

Meeting Notes:

• Discussed specific comments from the VICTR application and ways to address them

Recommendations:

• If considering a control group, a 2:1 ratio could work
• Will want to consider using Wilcoxon tests (or other non-parametric tests), since data is likely not normal