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.

Notes (2022)

2022 November 30

Gloria Nashed (Jo Ellen Wilson), Critical Care

Returning for another clinic after 11/9/22. Mentor confirmed.

  • Meeting Notes:
    • Follow up from 11/9/22 clinic
    • No inherent reason for a certain population receiving antipsychotics or not
    • Likely to be a small number of withdrawn patients
    • Likely to be a lot of missing data for catatonia assessments, which would make the outcome (catatonia free days) hard to compute
      • Subjects need at least one catatonia assessment to be eligible

  • Recommendations:
    • Unlikely to need a propensity score model, but do need to show a descriptive table providing support for this
    • If there is a lot of withdrawn subjects (5% would be considered a lot), it would be good to run a Cox proportional hazards model to account for censoring
      • This could just be a sensitivity analysis
    • For missing data for catatonia assessments, can calculate the proportion of days they are catatonia free and multiply by 14
      • Need to know if they have an assessment each day or not
    • For missing data in covariates, can use multiple imputation
    • No need to use robust standard errors for the clusters since there are only 4
      • Can just adjust for the parent study instead
    • Recommended to apply for a VICTR voucher

2022 November 16

Bo Stubblefield, Emergency Medicine

Returning for another clinic after 10/26/22. The investigator has some questions about imputation.

  • Meeting Notes:
    • Patients with low-risk venous thromboembolism (VTE) discharged from ED on Apixban
    • Main question: Did patients with bleeding risk assessed by retrospectively calculated by REITE score have increase bleeding events?
    • Different risk score that looks at mortality is what determines if they should be discharged or not (PESI)
      • Bleeding is not included in this assessment

  • Recommendations:
    • For missing data in the exposure variable, can use multiple imputation
      • Can include any clinical variables and demographic variables in the imputation model
    • For high rate of missing data in the outcome, do not suggest imputation
      • Not observing event -> recorded as no event is one way to do this
    • For the individual outcomes with small event rates, can use Fisher’s Exact test
      • Can use chi-square for larger event rates
    • For the composite outcome, can use a chi-square
    • For number of events = 100, this is a large enough sample size to build a model

2022 November 9

Gloria Nashed (Jo Ellen Wilson), Critical Care

Catatonia is often thought of in the psychiatric setting, but it can occur in a variety of critical care settings, with or without precedent psychiatric illness and may or may not be accompanied by other neuropsychiatric conditions such as delirium. It has a high mortality rate of up to 20% if left untreated (Fink M, Taylor M. 2006) and also substantially raises risk of several medical complications including immobilization, pulmonary embolism, autonomic instability, etc. (McCall et al. 1995). There is concern that although patients with catatonia and concurrent psychiatric symptoms may benefit from the administration of antipsychotics, there is the possibility that the antipsychotic medication may actually make catatonia worse or precipitate catatonic symptoms (Bahro et al., 1999; Segal, 2001). There are a few case studies written in the past which have attempted to investigate what is called antipsychotic-associated catatonic symptoms however the relationship is still not fully understood. The purpose of this study is to understand how common intensive care practices, particularly the use of antipsychotic medications commonly used in the ICU, may or may not induce catatonia. I am partnering with my mentor to do this project. We are wanting to use data already collected from the Delirium and Catatonia (DeCat) Prospective Cohort database and receive feedback on our statistical analysis plan that we will be submitting to VICTR for a biostatistics voucher. Mentor confirmed.

  • Meeting Notes:
    • Hypothesize that adults given any antipsychotics will have less catatonia-free days
    • Follow adults for 14 days -> collect delirium and catatonia status daily, but some subjects with withdrawal or be discharged before the 14-day period is up
    • Primary endpoint: catatonia-free days
      • If a subject dies, these days will count against catatonia-free days
  • Recommendations:
    • Need to consider if there is an inherent reason for why a certain population received antipsychotics or not
      • If there is, a propensity score model may be necessary
    • For missing outcome data, can come up with rules for imputation (ex. if catatonia is present on day 3 and 5, but day 4 is missing, can impute catatonia for day 4)
    • Otherwise, can run 2 sensitivity analyses:
      • Complete case vs some sort of imputation for catatonia outcome
      • Cox proportional hazards model to handle censored subjects
    • For the binary outcome, can plot partial effects plot
    • Can apply for a VICTR voucher, but recommended to come back to a clinic first to discuss the recommendations from today

2022 October 26

Jennifer Connell (Matthew Kynes), Anesthesiology - Global Health

We are hoping to do a feasibility study on the use of iron supplementation to correct pre-operative anemia in a low-income country. Our primary outcome is the increase in hemoglobin, secondary outcomes include transfusion rates, hospital/ICU admission, and LOS. There are not good data on baseline rates of pre-operative anemia or transfusion rates in low income countries. Therefore, this is a pilot study to inform a future, larger study determining the effectiveness of iron to reduce transfusions in this setting. Mentor confirmed.

  • Meeting Notes:
    • No show

Bo Stubblefield, Emergency Medicine

Need guidance on how to calculate sens, spec, PPV, NPV and ROC curve for a NLP algorithm that identifies pulmonary embolism from radiology reports in the adult ED. Our positive NLP hit rate for the last year (Oct 1 2021- sept 30 2022) was 272 cases. Our denominator of All CT imaging studies performed for PE was 2977.

  • Meeting Notes:
    • Using a tool to look over radiology reports for adults in the ED with chest scans to determine if they had a PE (yes/no)
    • 1 year of data currently collected with 2 adjudicators going through the reports
    • Want to calculate the validity of the tool

  • Recommendations:
    • Need to define the cohort
      • If just looking at the at-risk cohort (patients with symptoms of chest pain or shortness of breath), then sensitivity, specificity, NPV, and PPV can all be calculated
    • Could be interesting to look at asymptomatic patients with scans to determine if there are other reasons for PEs
    • For amount of data collection, 1 year could be better than 6 months to account for any seasonal effects
    • Can compare the two adjudicators in R

2022 September 21

Meredith Diaz (Julie Pingel, Jessica Anderson), Pediatric Pharmacy

Retrospective chart review to evaluate the safety of propofol use for non-procedural sedation in the PICU and PCICU. Mentor confirmed.
  • Meeting Notes:
    • Study group: all patients younger than 18 admitted to the PICU or PCICU who required mechanical ventilation and got propofol for non-procedural continuous sedation
    • No control group

  • Recommendations:
    • Without a control group, a decision about the use of propofol cannot be made
    • Can be a descriptive study
    • To say a change in opioid use before and after propofol is due to propofol, an assumption must be made that nothing else could possibly explain this change

Rachel McCaffrey, Surgery/Surgical Oncology

Retrospective review of all gender affirming top surgeries performed from 2019 to date to identify incidental pathologic findings and subsequent treatment outcomes. Goal to compare to a cohort of age matched reduction mammoplasty patients for risk factors associated with breast cancer or high risk lesion incidence and to determine if some transgender patients should be offered additional screening prior to surgery. Wish to discuss number of cases needed to detect a difference between gender surgery and mammoplasty patients in terms of risk of incidental pathology findings. Want to ensure we are not underpowered and that if indeed we detect no difference between populations that "no difference" is indeed true.

  • Meeting Notes:
    • Two groups: transgender top surgery and reduction mammoplasty
    • Retrospective chart review to identify high-risk cancer
    • Previous studies show about 1% incidence rate in transgender top surgery. Expect similar rate for reduction mammoplasty

  • Recommendations:
    • Potentially start with a case control study first
      • Can come back to a clinic to discuss this study design
    • Eventually do risk prediction once a lot more data has been collected
      • Would need a much larger sample size (Number of events / 15 = how many risk factors can be included)
    • As data is being collected, pay attention to the incidence rate to confirm if it is what is expected

2022 September 7

Michael Topf, Otolaryngology - Head and Neck Surgery

Design of a study that will examine surgeon, pathologist, radiation oncologist, radiologist marking of surgical margins on 3D scans. There has been some literature to support that surgeons often have significant difficulty in re-locating positive margins in head and neck cancer surgery. We are trying to quantify how far off surgeons are when marking margins using existing 3D scans of specimens.

  • Meeting Notes:
    • Research question: how accurate are head and neck cancer surgeons when relocating positive margins?
    • Today’s main focus is on sample size

  • Recommendations:
    • Can use margin of error formula to determine sample size
    • Margin of error formula:
      • ((1.96*5.7)/MOE)^2
        • z = 1.96; omega = 5.7; margin of error = 2 -> N = 31
    • Need to be confident that measurements are independent of each other
      • If not, could take the average of how far off each participant is
      • Would need N = 31 participants, regardless of how many measurements for each one
      • If taking this approach, this would be quantifying the difference between each participant instead of each measurement
      • Using the average, the SD would narrow as the number of measurements per participant increases

2022 August 17

Jessica Miller (Ashish Patel, Meri Johnson), Radioligy

This project is a revision of a previously submitted project. The goal is to look at ED head imaging in minor trauma, headache, and altered mental status and determine if ordering patterns follow appropriateness criteria. The ED EPIC note and additional clinical data from the encounter will be used to determine if guidelines are followed. This will be investigate before and after an intervention with training to ED providers on appropriateness criteria using the Quiz Time platform. I would like to discuss the sample size and study duration, and the basic study design to ensure I am not overlooking something major. Mentor confirmed.

2022 July 27

Shelby Meier (Jana Shirey-Rice), VICTR - Drug Repurposing Team

Novel analyses on existing clinical trial data (HEADWAY-DLB) Proposed analysis #1: Reanalyze data to look at novel symptomology of interest to look for hints of efficacy Proposed analysis #2: Redefine patient eligibility based on 5HT-6 target related symptoms/presentation, then look at outcomes (both previous reported + novel symptomology) Proposed Analysis #3: Identify the clinical profile of “drug responders” based on previous or proposed endpoints; compare to precise phenotypes developed from chart reviews Questions: 1) Are these analyses feasible and appropriate? 2) Would there be better analyses to do instead of the ones proposed? Mentor confirmed.

  • Meeting Notes:
    • Data come from a previous clinical trial (everything is collected)
    • Symptoms of interest include hallucinations, autonomic dysfunction, REM sleep disorder, and loss of smell
  • Recommendations:
    • Can use differential treatment effects for subgroup analyses
      • Fit the model that was planned for the study and then include the interaction term between the drug and whatever the subgrouping variable is
    • Recommended to apply for a VICTR voucher
      • To keep this project from getting too big for a voucher, list out everything that you want to test (hypotheses for outcomes and effect modifiers)

Kristina Marie Niehoff, Vanderbilt Home Care Services

Biostats assistance for "Assessment of apixban for venous thromboembolism prophylaxis in Vanderbilt’s COVID to Home Program" Completed VICTR design studio and outcome was to book time with a biostats expert.

  • Meeting Notes:
    • Some limitations: small number of outcomes
    • Primary outcomes: frequency and type of bleeding events
    • Secondary outcomes: frequency of ER visits and hospitalizations
    • The cohort includes inpatients who have been discharged

  • Recommendations:
    • Make it clear what cohort is being looked out. This will distinguish this study from other studies on the same topic.
    • This may already be approved for a voucher – following up with Chris to check on this. If it is not already approved, it is recommended to apply for a voucher.

2022 July 20

Isaura Diaz, Pediatric Critical Care

We devised a national survey to evaluate the number of PICU programs in our Nation have a cardiac simulation curriculum in place. Out of 72 programs we had 32 programs were represented and 16 of these programs reported a cardiac simulation curriculum that was in place. We are now writing up a manuscipt regarding our findings and would like to confirm the statistical portion of our findings.

  • Meeting Notes:
    • 72 programs were sent a survey; 32 responded
      • Of the 32 responses, 16 programs had a simulation curriculum
      • N = 156 participants (fellows) from these 32 programs responded
    • The survey asks about comfort level during different procedures using a Likert scale to answer
  • Recommendations:
    • Can use a mixed effects model and get an odds ratio
      • This will be more meaningful than the results from a Mann-Whitney U test
    • In Table 1, look at the data at the PCICU level
    • Can apply for a VICTR voucher for this part of the study
      • For the second phase, a second voucher may be necessary (clarifying with Heather if this is allowed)
    • Attend another clinic to discuss the power question for the other study

Olatundun Ladele (Izabela Galdyn), Department of Plastic Surgery

The goal of this project is to use the NASS database to investigate preoperative risk factors and the complications that present after inpatient treatment of mandibular fracture and estimate the best complication rate. Before conclusion of data analysis, my hypothesis is that there will be an increase in complications following treatment in patients with preoperative risk factors such as diabetes, obesity, and tobacco use. Questions: Running multivariate correlations in order to assess if there are preoperative risk factors for the complications. Mentor confirmed.

  • Meeting Notes:
    • ICD codes were used to determine who was treated for mandibular fractures
    • Complications were chosen based on literature review
    • Physicians reported complications after surgery
  • Recommendations:
    • Run univariate analyses to look at the risk factors/exposures one by one with the primary outcome
      • Present odds ratios and point estimates + 95% confidence intervals
    • Run a multivariate analysis including all the risk factors/exposures
      • Can base this on clinical experience + univariate analysis results
      • Or could run LASSO to determine factors to include in the model

2022 July 13

Shiayin Yang, Otolaryngology

We are investigating the effect that reconstruction on the nose after cancer removal has on breathing. We would like to determine if certain nasal areas, depth of defect, and type of reconstruction have increased likelihood of nasal obstruction after Mohs surgery.

  • Meeting Notes:
    • Want to look at how different factors with surgery affect nasal breathing
      • Important factors: location, type of surgery, layers, cartilage (yes/no)
    • Have data overtime: baseline, 1 month post-surgery, 3 months post-surgery, 6 months post-surgery, 1 year post-surgery

  • Recommendations:
    • Mixed effects model with participant as a random effect
      • Include a time by surgery interaction
    • Include variables that you think would be different between a good and bad breathing outcome
    • Will need to consider degrees of freedom and sample size
    • With around 20% of the sample size having a bad breathing outcome, will likely need around N = 300-400 overall.
    • Recommended to apply for a VICTR voucher

2022 June 29

Ankush Ratwani (Fabien Maldonado), Pulmonary and Critical Care

Question for Chris Lindsell, already aware of the project. Question relating to exclusion after randomization and design for our study. Mentor confirmed.

  • Meeting Notes:
    • Pragmatic multi-centered randomized trial
    • Main question: how to account for exclusions after randomization
  • Recommendations:
    • Could include a post-randomization exclusion if only those with a malignancy are included after the envelope is opened
      • Could maybe reuse these envelopes from people who are excluded post-randomization (would need to consider how many people this would be)
    • Could issue the randomization envelope right when going into surgery
    • Power is based on those that get included

2022 June 22

Raymond Zhou (Louise Mawn), Ophthalmology and Visual Sciences

IIH is a growing health concern affecting obese young women, with risk factors and treatments that are under continual study. We want to identify what risk factors and treatments make severe vision loss, measured by visual field mean defect (VFMD) and visual acuity. in patients with IIH more or less likely. We intend to do so by studying all patients with IIH that were previously treated at VUMC. We intend to collect data on the presence or absence of various risk factors, including obesity, race, and gender in these patients. We also intend to collect data on the provision of various invasive treatments that release pressure and divert fluid pushing on the optic nerve. Our primary statistical analysis will utilize a multivariable linear regression model with the outcomes being the continuous variables VFMD and Visual Acuity. Exactly as suggested by VICTR, this will increase the power of our study. Secondary analyses will be done using the dichotomous variable +/-decrease in visual acuity, and +/- decrease in VFMD. While previous studies have studied these dichotomous variables, this has several drawbacks including loss of efficiency and potential for ceiling/floor effects. We appreciate the question and recommendation that was provided by VICTR. We would like to further obtain feedback regarding our proposed statistical analyses. Additionally, we would like to potentially explore the potential need to obtain additional biostatistical support, as our co-collaborators will be beginning residency soon. The Department of Ophthalmology has a designated institutional partnership with Dr. Chen, who we would love to work with. It is my understanding that, to collaborate with Dr. Chen, projects must be discussed at a Biostats clinic. Mentor confirmed.

  • Meeting Notes:
    • Goal: Want to identify risk factors and treatments that make severe vision loss (measured by visual field mean defect - VFMD and visual acuity)
    • Some potential challenges:
      • Is the cohort balanced at baseline?
      • How to account for multiple visits before surgery

  • Recommendations:
    • Can use a propensity score method
      • Get the propensity for surgery among the cohort, and then adjust for the propensity score or using propensity score matching
      • In the propensity score model, include factors that are associated with the physician’s decision to treat (will be a combo of patient and physician factors)
    • Check with your collaboration first; if you cannot work with them, this project is appropriate for a VICTR voucher

Shaan Setia (Ryan Hsi), Urology

ITS analysis. Mentor confirmed.

  • Meeting Notes:
    • Looking at outcomes from before the new outpatient center opened compared to after
    • Hypothesize that outcomes are better with the new center
    • Includes data from 6 months before the center opened and 6 months after the center opened

  • Recommendations:
    • Use an interrupted time series analysis
      • Doing this instead of a t-test allows us to see if outcomes are just getting better over time, or if there is a difference before and after the opening of the center
      • Include an interaction between time and the pre/post indicator variable
    • Apply for a VICTR voucher

2022 April 20

Kathy Mendieta (Katie Boyle), Pediatric Critical Care

Our QI project entails the use of a CPR Coach during resuscitation events in both the medical and cardiac ICU. Since implementing this new role (April 2021) we have collected data via survey following a resuscitation event. We are in the process of reviewing objective data (blood pressure, end-tidal; if compressors switched every 2 minutes, and presence of equipment required for high-quality CPR) and comparing codes with a CPR coach versus without. Mentor confirmed.
  • Meeting Notes:
    • Main goal: after implementation of the new role, want to compare codes with and without a CPR coach (compare vital signs and use of new tools)
    • N = 46 codes collected in the survey currently (around 60% with a CPR coach)
      • All after the implementation of the CPR coach
  • Recommendations:
    • Recognizing this is an observational study, can go ahead with the current number of cases.
    • Can estimate the magnitude of the difference between the two groups
    • Need to recognize that it is unknown who got a coach and why
    • Recommended to apply for a VICTR voucher?

Allison Weatherly (Kristina Betters), Pediatric Critical Care

We want to look at the relationship between hyperoxia on bypass and delirium outcomes in patients in the pediatric cardiac ICU via a retrospective chart review. Delirium is prevalent in this population (seen in up 60%+ of patients) and so we're wondering how many patients we would need to power this study. Additionally, we have been discussing how to measure/analyze hyperoxia. We are thinking about looking at time spent at a given paO2 vs change in NIRS while on bypass (we have minute to minute data on NIRS) and so discussions about what data to pull from the bypass records and how we would potentially analyze this. Other hyperoxia studies that we have looked at, look at AUC vs separate paO2 into quartiles. I think that this particularly needs to be flushed out as it will affect data collection significantly. Mentor confirmed.
  • Meeting Notes:
    • Multiple ways to define delirium
    • Is a slightly hyperoxia over a long time and large hyperoxia in a short time the same thing?
    • Around 40-60% in this group will have delirium
  • Recommendations:
    • Change the research question to: Do different measures of hyperoxia have different associations with delirium?
    • Write down the different ways to measure hyperoxia
    • Will be an exploratory study with multiple exposures
    • If about 50% of the group has delirium, this will maximize the power
    • Recommended to apply for a VICTR voucher

2022 March 2

Chiara Foster (Kristina Betters), Pediatric Critical Care

Cost analysis of a sedation weaning protocol. Assessing how different the two groups are and comparing costs with median and interquartile ranges as well as discussing if secondary analyses are indicated. Mentor confirmed.
  • Meeting Notes:
    • Costs are usually not distributing normally
    • Costs in this study are purely pharmaceutical (no hospital costs)
    • Three different drugs: some patients are on all 3, but some are only on 1 or 2
      • Have a risk score for each drug, but not an overall risk score
  • Recommendations:
    • To statistically compare the 2 groups, could compare the medians
      • Mann-Whitely U test (does not adjust for risk)
      • To adjust for previous exposure, could use a proportional odds model or a Cox model, adjusting for risk
    • Can use a chi-square test to compare the risk between the two groups
    • For the VICTR voucher, Chris will help with the appeal process

Christina Boncyk, Anesthesiology

The primary aim of this observational study is to utilize the electronic health record to identify and validate environmental and structural SDOH and to determine the association between these SDOH factors and ICU-free days over the first 28 days following admission in critically ill patients with COVID-19. We believe that there will be a correlation between the identified markers of SDOH and ICU free days at 28-days following admission.

We plan to use logistic regression adjusting for the Charlson comorbidity index (CCI), the Elixhauser-based comorbidity summary, and the van Walraven summary score. Secondary outcomes will include ventilator-free days over 28-days, 28-day in-hospital mortality, duration of hospitalization, and discharge location. We will perform linear regression modelling adjusting for the same confounders of CCI, Elixhauser comorbidities, and van Walraven summary score.
  • Meeting Notes:
    • Hypothesis: SDOH is associated with in-hospital outcomes among critically ill adults diagnosed with COVID-19
    • Primary outcome: ICU-free days over first 28 days
    • Secondary outcomes: ventilator-free days, 28-day in-hospital mortality, hospital length of stay, discharge location
    • Overall goal: look at different SDOH associations with the outcomes
  • Recommendations:
    • Can test the SDOH one by one
    • A VICTR voucher may be appropriate
      • First need to identify variables to include
      • Chris can help with the SAP
    • Generating and validating a predictive model is likely too big for a VICTR voucher

Current planning and questions we are hoping to answer would be discussion on how study plan could benefit from a biostatistician with an expertise in geospatial/geographic data, what we need to include while constructing our data output, and other experience within this type of analysis.

2022 February 16

Bo Stubblefield, Emergency Medicine

Analysis of bleeding outcomes in a subset of patient with acute pulmonary embolism and high risk features.
  • Meeting Notes:
    • Main question: Do higher risk patients have worse bleeding outcomes?
    • Primary outcome: could make a composite outcome (unscheduled visit for bleeding, rehospitalization for bleeding, transfusion) – all within 30 days
    • Potential secondary outcomes: complications (renal failure, liver failure, stroke, MI); rehospitalization for PE of DVT
  • Recommendations:
    • Any conclusions can only be applied to those who are discharged home
    • Seems more like an exploratory risk factor analysis - explore the prevalence of each exposure
    • Look into the derivation of the PESI score and find all their candidate risk factors
      • Can look at ones that were considered and ones that were not, but it will be important to note

2022 January 19

Montana (Lori) Fleenor (Joanna Stollings), Pharmacy

This study is a retrospective cohort study looking at the effects of tocilizumab on prevalence of delirium and coma in COVID-19 patients. Sample size will consist of a control vs. tocilizumab group (anticipate ~120-150 pts total). We aim to characterize the incidence and duration of delirium/coma in these patients. Independent variable: control vs. tocilizumab group. Dependent variable: Days alive without delirium/coma (denominator of 21 days). At least one instance of delirium by a positive CAM-ICU will be considered delirious for that day. At least one instance of coma by a RASS -4/-5 will be considered comatose for that specific day. Also need to incorporate death into the equation because this is looking at days alive without delirium or coma. Covariates: Age, sex, CRP, Charlson Comorbidity Index, SOFA, and vent status. Proposed statistical model: Proportional odds logistic regression. Questions: 1)What is your opinion on adding these other covariates? -sepsis, steroids use, delirium, total daily dose of analgesics and sedatives (benzodiazepine, propofol, opioids, dexmedetomidine) 2)Will it be beneficial to also consider precision variables, that only affect dependent variables? 3)Do we need additional statistical models? 4)Should we look at ICU, hospital length of stay, and 90 day mortality? VICTR Biostatistics voucher. Mentor confirmed.
  • Meeting Notes:
    • Looking at delirium outcomes before and after TOCI was used as the standard practice for COVID patients
    • Proposed model is a proportional odds logistic regression
    • Primary outcome: days alive in the ICU without delirium/coma within 21 days of receiving TOCI (believe TOCI effects will be gone by 21 days)
  • Recommendations:
    • Might consider an interrupted time series model
    • Clarify how to count days alive for subjects who die in less than 21 days
    • Include any covariates that would be clinically would be a confounder of delirium
    • Include analyses for secondary outcomes of interest
    • Recommended to apply for a VICTR voucher

2022 January 12

Margaret Barton (Donald Arnold), Pediatric Emergency Medicine

Children discharged from emergency departments (EDs) are at risk to return to the ED. Those who return within 48-72 hours of their visit may represent gaps in patient care including proper assessment, treatment, or follow-up instructions. Additionally, there are likely demographic factors that influence rate of ED return visits. Identifying areas of possible intervention to reduce ED return visits may help hospitals take care of patients better, reduce ED crowding, and reduce costs. The aim of this study is to describe the incidence of and identify risk factors for 48 and 72-hour ED return visits among this patient population. We will obtain demographic (age, language, zip-code, PCP status, insurance) and visit characteristic data (chief complaint, triage acuity, diagnosis, reason for return, medication/imaging received during visit, medication prescribed after visit, follow-up plan in place). We will conduct a single-center retrospective study at our children’s hospital of all 48 and 72-hour return visits during a 6-year period (January 2017 to December 2022). Questions: How to organize demographic data of both 48h and 72h return visits (some patients will overlap)? We would like to see if certain zip codes correspond to different rates of return. In particular we are interested to see whether zip-codes with significantly different median incomes vary in their rate of return. How would you recommend comparing this data? Should we use excel or redcap? Overall, we would like help in organizing our data collection so it is easy to analyze. VICTR Biostatistics voucher. Mentor confirmed.
  • Meeting Notes:
    • Overall objective is to describe the incidence of 48 and 72-hour return ED visits and to identify risk factors (demographic + clinical data from first visit)
    • Most patients that come to Vanderbilt Children’s ED would return here for a return visit
  • Recommendations:
    • Need to work on focusing the research question
    • One issue to consider: patients come back for different reasons, and this may be hard to tease out, but could look at comparing the reason for why they left the ER in the first place
    • Consider today’s discussion and then come back to another Wednesday clinic with Chris
Topic revision: r1 - 16 Dec 2022, DalePlummer
 

This site is powered by FoswikiCopyright © 2013-2022 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback