Katharine Alison Miller
Assistant Director of Data Visualization Program, Med/Quantitative Sciences Unit
All Publications
-
Value of diversity characteristics in predictive modeling: ACS screening as a case study.
NPJ cardiovascular health
2025; 2 (1): 52
Abstract
We sought to improve the subgroup performance variability of a model that identifies arriving ED patients at high risk for ACS, to receive an ECG within 10 minutes of arrival, to detect STEMI. We compared a Base Model using age, sex, and chief complaint alone to (1) an Interactions Model adding interactions among the 3 variables, and (2) a Diversity-Sensitive Model including race, ethnicity, language, as well as identity interactions. We quantified human performance and combined it with each of the 3 models simulating use as practice augmenting AI predictions. With sensitivity as our primary outcome, we found humans at 72.8% were bested by the Diversity-Sensitive Model at 82.8%, and by the human-augmented Diversity-Sensitive Model at 91.3%, improving ACS predictions in all demographic subgroups. However, there was residual variation among subgroups (range of sensitivity: 62%-98%). Given risk distribution differences, subgroup-specific ECG-testing thresholds may further equitize ACS prediction performance.
View details for DOI 10.1038/s44325-025-00088-0
View details for PubMedID 41133128
View details for PubMedCentralID PMC12540186
-
In vitro to in vivo translation of artificial intelligence for clinical use: screening for acute coronary syndrome to identify ST-elevation myocardial infarction.
Journal of the American Medical Informatics Association : JAMIA
2025
Abstract
The integration of predictive models into live clinical care requires scientific testing before implementation to ensure patient safety. We built and technically implemented a model that predicts which patients require an electrocardiogram (ECG) to screen for heart attacks within 10 minutes of their arrival to the Emergency Department. We developed a structured framework for the in vitro to in vivo translation of the model through implementation as clinical decision support (CDS).The CDS ran as a silent pilot for 2 months. We conducted (1) a Technical Component Analysis to ensure each part of the CDS coding functioned as planned, and (2) a Technical Fidelity Analysis to ensure agreement between the CDS's in vivo and the model's in vitro screening decisions.The Technical Component Analysis indicated several small coding errors in CDS components that were addressed. During this period, the CDS processed 18 335 patient encounters. CDS fidelity to the model reflected raw agreement of 95.5% (CI, 95.2%-95.9%) and kappa of 87.6% (CI, 86.7%-88.6%). Additional coding errors were identified and were corrected.Our structured framework for the in vitro to in vivo translation of our predictive model uncovered ways to improve performance in vivo and the validity of risk assessment decisions. Testing predictive models on live care data and accompanying analyses is necessary to safely implement a predictive model for clinical use.We developed a method for the translation of our model from in vitro to in vivo that can be utilized with other applications of predictive modeling in healthcare.
View details for DOI 10.1093/jamia/ocaf101
View details for PubMedID 40576204
-
Emotional Reactivity and Regulation Relate to Surgical Treatment Decision Making Among Newly Diagnosed Women With Breast Cancer.
Cancer medicine
2024; 13 (23): e70357
Abstract
Despite bilateral mastectomy (BLM) for early-stage breast cancer (BC) showing no survival benefit and increased risk compared to breast conserving surgery, some patients still choose this treatment. This study examined whether emotion reactivity and regulation influence treatment decision making among newly diagnosed women with breast cancer.Cross-sectional survey data were analyzed as part of a larger study. Measures included the Contralateral Prophylactic Mastectomy (CPM) survey, PROMIS Anxiety scale, and Emotion Regulation Questionnaire (ERQ) Cognitive Reappraisal and Emotional Suppression subscales. Primary analysis included a logistic regression model predicting treatment choice (BLM vs. non-BLM).Participants (N = 137) with unilateral BC (Stages 0-III) were divided between BLM (n = 66) versus breast conserving surgery (i.e., non-BLM, n = 71) treatment groups. Compared to the non-BLM group, the BLM group was younger, more likely to be partnered, and had a higher household income. Women with high levels of BC-specific worry were 3.6 times more likely to choose BLM compared to women with low levels of worry (OR = 3.09, 95% CI: 1.07-0.8.93). Those who used cognitive reappraisal were 10% less likely to choose BLM compared to women who did not use cognitive reappraisal (OR = 0.90, 95% CI: 0.82-0.99). There were no group differences in levels of generalized anxiety (OR = 0.93, 95% CI: 0.87-0.99) or emotional suppression (OR = 1.02, 95% CI: 0.90-1.16).Findings suggest the choice of BLM may be due, in part, to negative emotional experiences after a BC diagnosis and lesser use of reappraisal to reframe cancer-related fears. These may be important targets of intervention to support women making BC treatment decisions.
View details for DOI 10.1002/cam4.70357
View details for PubMedID 39653041
-
A Perioperative Quality Improvement Program for Cesarean Delivery in Ethiopia: A Stepped-Wedge Cluster Randomized Clinical Trial.
JAMA network open
2024; 7 (8): e2428910
Abstract
Infections and complications following cesarean delivery are a significant source of maternal mortality in Ethiopia.To study the effectiveness of a program to strengthen compliance with perioperative standards and reduce postoperative complications following cesarean delivery.This stepped-wedge cluster randomized clinical trial included patients undergoing cesarean delivery from August 24, 2021, to January 31, 2023, at 9 hospitals organized into 5 clusters in Ethiopia.Clean Cut, a multimodal surgical quality improvement program that includes process-mapping 6 perioperative standards and creating site-specific, systems-level improvements. The control period was the period before implementation of the intervention.The primary end point was surgical site infection rate, and secondary end points were maternal mortality and perinatal mortality and a composite outcome of infections and both mortality outcomes. All were assessed at 30 days postoperatively in the intervention and control groups, adjusting for clustering and demographics. Compliance with standards and the relationship between compliance and outcomes were also compared between the 2 arms.Among 9755 women undergoing cesarean delivery, 5099 deliveries (52.3%) occurred during the control period (2722 emergency cases [53.4%]) and 4656 (47.7%) during the intervention period (2346 emergency cases [50.4%]). Mean (SD) patient age was 27.04 (0.05) years. Thirty-day follow-up was completed for 5153 patients (52.8%). No significant reduction in infection rates was detected after the intervention (OR, 0.84; 95% CI, 0.55-1.27; P = .40). Intraoperative infection prevention standards improved significantly in the intervention arm vs control arm for compliance with at least 5 of the 6 standards (odds ratio [OR], 2.95; 95% CI, 2.40-3.62; P < .001). Regardless of trial arm, high compliance was associated with reduced odds of maternal (OR, 0.32; 95% CI, 0.11-0.93; P = .04) and perinatal (OR, 0.64; 95% CI, 0.47-0.89; P = .008) mortality.In this stepped-wedge cluster randomized clinical trial of patients undergoing cesarean delivery, no significant reductions in surgical site infections were observed. However, compliance with perioperative standards improved following the intervention.ClinicalTrials.gov Identifier: NCT04812522; Pan-African Clinical Trials Registry Identifier: PACTR202108717887402.
View details for DOI 10.1001/jamanetworkopen.2024.28910
View details for PubMedID 39163043
-
Travel-Time barriers to specialized cancer care for adolescents and young adults with acute leukemia.
JNCI cancer spectrum
2024
Abstract
BACKGROUND: Prior studies demonstrate that 20-50% of adolescents and young adults (AYA, age 15-39years) with acute lymphoblastic leukemia (ALL) receive care at specialty cancer centers (SCC); yet a significant survival benefit has been observed for patients at these sites. Our objective was to identify patients at risk of severe geographic barriers to SCC-level care.METHODS: We used data from the North American Association of Central Cancer Registries Cancer in North America database to identify AYA ALL patients diagnosed between 2004-2016 across 43U.S. states. We calculated driving distance and travel time from counties where participants lived to the closest SCC sites. We then used multivariable logistic regression models to examine the relationship between sociodemographic characteristics of counties where AYA ALLs resided and the need to travel >1hour to obtain care at an SCC.RESULTS: Among 11,813 AYA ALL patients, 43.4% were 25-39years old, 65.5% were male, 32.9% were Hispanic, and 28.7% had public insurance. We found 23.6% of AYA ALL patients from 60.8% of included U.S. counties would be required to travel >1hour one-way to access an SCC. Multivariable models demonstrate that patients living in counties that are non-metropolitan, with lower levels of educational attainment, with higher income inequality, lower internet access, located in primary care physician shortage areas and with fewer hospitals providing chemotherapy services are more likely to travel >1hour to access an SCC.CONCLUSIONS: Substantial travel-related barriers exist to accessing care at SCCs across the U.S, particularly for patients living in areas with greater concentrations of historically marginalized communities.
View details for DOI 10.1093/jncics/pkae046
View details for PubMedID 38845074
-
A phase II study of plerixafor combined with whole brain radiation therapy (WBRT) for patients with newly diagnosed glioblastoma
LIPPINCOTT WILLIAMS & WILKINS. 2024
View details for Web of Science ID 001275557400433
-
Impact of Specialized Treatment Setting on Survival in Adolescent and Young Adult ALL.
JCO oncology practice
2023: OP2300373
Abstract
Unlike children with ALL who receive cancer care primarily at specialized cancer centers (SCCs; National Cancer Institute and/or Children's Oncology Group centers), adolescents and young adults (AYAs; 15-39 years) receive care in a variety of settings. Using population-based data, we describe where AYAs with ALL receive treatment and determine associations with overall survival (OS).Data from the 2004 to 2018 California (CA, n = 2,283), New York (NY, n = 795), and Texas (TX, n = 955) state cancer registries were used to identify treatment setting of AYAs with newly diagnosed ALL. Multivariable Cox proportional hazards regression models evaluated associations with OS.Seventy percent were older than 18 years, and 65% were male. A majority in CA (63%) and TX (64%) were Hispanic while most in NY were non-Hispanic White (50%). Treatment at an SCC occurred in 48.2% (CA), 44.4% (NY), and 19.5% (TX). Across states, AYAs who were older or uninsured were less likely to receive treatment at an SCC. Treatment at an SCC was associated with superior OS in CA (hazard ratio [HR], 0.73; 95% CI, 0.63 to 0.85) and TX (HR, 0.61; 95% CI, 0.45 to 0.83); a nonsignificant association was seen in NY (HR, 0.83; 95% CI, 0.64 to 1.08).Only 20%-50% of AYA patients with ALL received frontline treatment at SCCs. Treatment of ALL at an SCC was associated with superior survival, highlighting the importance of policy efforts to improve access and reduce inequities in AYA ALL care.
View details for DOI 10.1200/OP.23.00373
View details for PubMedID 37890123
-
Race and Ethnicity and Prehospital Use of Opioid or Ketamine Analgesia in Acute Traumatic Injury.
JAMA network open
2023; 6 (10): e2338070
Abstract
Racial and ethnic disparities in pain management have been characterized in many hospital-based settings. Painful traumatic injuries are a common reason for 911 activations of the EMS (emergency medical services) system.To evaluate whether, among patients treated by EMS with traumatic injuries, race and ethnicity are associated with either disparate recording of pain scores or disparate administration of analgesia when a high pain score is recorded.This cohort study included interactions from 2019 to 2021 for US patients ages 14 to 99 years who had experienced painful acute traumatic injuries and were treated and transported by an advanced life support unit following the activation of the 911 EMS system. The data were analyzed in January 2023.Acute painful traumatic injuries including burns.Outcomes were the recording of a pain score and the administration of a nonoral opioid or ketamine.The study cohort included 4 781 396 EMS activations for acute traumatic injury, with a median (IQR) patient age of 59 (35-78) years (2 497 053 female [52.2%]; 31 266 American Indian or Alaskan Native [0.7%]; 59 713 Asian [1.2%]; 742 931 Black [15.5%], 411 934 Hispanic or Latino [8.6%], 10 747 Native Hawaiian or other Pacific Islander [0.2%]; 2 764 499 White [57.8%]; 16 161 multiple races [0.3%]). The analysis showed that race and ethnicity was associated with the likelihood of having a pain score recorded. Compared with White patients, American Indian and Alaskan Native patients had the lowest adjusted odds ratio (AOR) of having a pain score recorded (AOR, 0.74; 95% CI, 0.71-0.76). Among patients for whom a high pain score was recorded (between 7 and 10 out of 10), Black patients were about half as likely to receive opioid or ketamine analgesia as White patients (AOR, 0.53; 95% CI, 0.52-0.54) despite having a pain score recorded almost as frequently as White patients.In this nationwide study of patients treated by EMS for acute traumatic injuries, patients from racial or ethnic minority groups were less likely to have a pain score recorded, with Native American and Alaskan Natives the least likely to have a pain score recorded. Among patients with a high pain score, patients from racial and ethnic minority groups were also significantly less likely to receive opioid or ketamine analgesia treatment, with Black patients having the lowest adjusted odds of receiving these treatments.
View details for DOI 10.1001/jamanetworkopen.2023.38070
View details for PubMedID 37847499
-
Changes in low-acuity patient volume in an emergency department after launching a walk-in clinic.
Journal of the American College of Emergency Physicians open
2023; 4 (4): e13011
Abstract
Unscheduled low-acuity care options are on the rise and are often expected to reduce emergency department (ED) visits. We opened an ED-staffed walk-in clinic (WIC) as an alternative care location for low-acuity patients at a time when ED visits exceeded facility capacity and the impending flu season was anticipated to increase visits further, and we assessed whether low-acuity ED patient visits decreased after opening the WIC.In this retrospective cohort study, we compared patient and clinical visit characteristics of the ED and WIC patients and conducted interrupted time-series analyses to quantify the impact of the WIC on low-acuity ED patient visit volume and the trend.There were 27,211 low-acuity ED visits (22.7% of total ED visits), and 7,058 patients seen in the WIC from February 26, 2018, to November 17, 2019. Low-acuity patient visits in the ED reduced significantly immediately after the WIC opened (P = 0.01). In the subsequent months, however, patient volume trended back to pre-WIC volumes such that there was no significant impact at 6, 9, or 12 months (P = 0.07). Had WIC patients been seen in the main ED, low-acuity volume would have been 27% of the total volume rather than the 22.7% that was observed.The WIC did not result in a sustained reduction in low-acuity patients in the main ED. However, it enabled emergency staff to see low-acuity patients in a lower resource setting during times when ED capacity was limited.
View details for DOI 10.1002/emp2.13011
View details for PubMedID 37484497
View details for PubMedCentralID PMC10361543
-
Maximizing Equity in Acute Coronary Syndrome Screening across Sociodemographic Characteristics of Patients.
Diagnostics (Basel, Switzerland)
2023; 13 (12)
Abstract
We compared four methods to screen emergency department (ED) patients for an early electrocardiogram (ECG) to diagnose ST-elevation myocardial infarction (STEMI) in a 5-year retrospective cohort through observed practice, objective application of screening protocol criteria, a predictive model, and a model augmenting human practice. We measured screening performance by sensitivity, missed acute coronary syndrome (ACS) and STEMI, and the number of ECGs required. Our cohort of 279,132 ED visits included 1397 patients who had a diagnosis of ACS. We found that screening by observed practice augmented with the model delivered the highest sensitivity for detecting ACS (92.9%, 95%CI: 91.4-94.2%) and showed little variation across sex, race, ethnicity, language, and age, demonstrating equity. Although it missed a few cases of ACS (7.6%) and STEMI (4.4%), it did require ECGs on an additional 11.1% of patients compared to current practice. Screening by protocol performed the worst, underdiagnosing young, Black, Native American, Alaskan or Hawaiian/Pacific Islander, and Hispanic patients. Thus, adding a predictive model to augment human practice improved the detection of ACS and STEMI and did so most equitably across the groups. Hence, combining human and model screening--rather than relying on either alone--may maximize ACS screening performance and equity.
View details for DOI 10.3390/diagnostics13122053
View details for PubMedID 37370948
View details for PubMedCentralID PMC10297640
-
Specialty cancer care and survival in AYA ALL: A US multi-state analysis.
LIPPINCOTT WILLIAMS & WILKINS. 2023
View details for Web of Science ID 001053772003380
-
Evaluation of an adaptive, multimodal intervention to reduce postoperative infections following cesarean delivery in Ethiopia: study protocol of the CLEAN-CS cluster-randomized stepped wedge interventional trial.
Trials
2022; 23 (1): 692
Abstract
BACKGROUND: We previously developed and pilot tested Clean Cut, a program to prevent postoperative infections by improving compliance with the WHO Surgical Safety Checklist (SSC) and strengthening adherence to infection control practices. This protocol describes the CheckList Expansion for Antisepsis and iNfection Control in Cesarean Section (CLEAN-CS) trial evaluating our program's ability to reduce infections following CS and other obstetric and gynecological operations in Ethiopia.METHODS/DESIGN: CLEAN-CS is a cluster-randomized stepped wedge interventional trial with five clusters (two hospitals per cluster). It aims to assess the impact of Clean Cut on six critical perioperative infection prevention standards including antiseptic practices, antibiotic administration, and routine SCC use. The trial involves baseline data collection followed by Clean Cut training and implementation in each cluster in randomized order. The intervention consists of (1) modifying and implementing the SSC to fit local practices, (2) process mapping each standard, (3) coupling data and processes with site-specific action plans for improvement, and (4) targeted training focused on process gaps. The primary outcome is 30-day CS infection rates; secondary outcomes include other patient-level complications and compliance with standards. Assuming baseline SSI incidence of 12%, an effect size of 25% absolute reduction, and the ability to recruit 80-90 patients per cluster per month, we require a sample of 8100 patients for significance. We will report our study according to CONSORT.DISCUSSION: A cluster-randomized stepped wedge design is well-suited for evaluating this type of surgical safety program. The targeted standards are not in doubt, yet compliance is frequently difficult. Solutions are available and may be recognized by individuals, but teams dedicated to improvement are often lacking. Clean Cut was successfully piloted but requires a more rigorous methodological assessment. We seek to understand the qualities, characteristics, and resources needed to implement the program, the magnitude of effect on processes and outcomes, and to what degree it can enhance compliance with care standards. Challenges include a fraught social and political environment, pandemic travel restrictions, and a limited budget.TRIAL REGISTRATION: ClinicalTrials.gov NCT04812522 (registered on March 23, 2021); Pan-African Clinical Trials Registry PACTR202108717887402 (registered on August 24, 2021).
View details for DOI 10.1186/s13063-022-06500-9
View details for PubMedID 35986400