Development and Validation of a 30-Day In-hospital Mortality Model Among Seriously Ill Transferred Patients: a Retrospective Cohort Study

Abstract

BACKGROUND: Predicting the risk of in-hospital mortality on admission is challenging but essential for risk stratification of patient outcomes and designing an appropriate plan-of-care, especially among transferred patients. OBJECTIVE: Develop a model that uses administrative and clinical data within 24 h of transfer to predict 30-day in-hospital mortality at an Academic Health Center (AHC). DESIGN: Retrospective cohort study. We used 30 putative variables in a multiple logistic regression model in the full data set (n = 10,389) to identify 20 candidate variables obtained from the electronic medical record (EMR) within 24 h of admission that were associated with 30-day in-hospital mortality (p < 0.05). These 20 variables were tested using multiple logistic regression and area under the curve (AUC)-receiver operating characteristics (ROC) analysis to identify an optimal risk threshold score in a randomly split derivation sample (n = 5194) which was then examined in the validation sample (n = 5195). PARTICIPANTS: Ten thousand three hundred eightynine patients greater than 18 years transferred to the Indiana University (IU)-Adult Academic Health Center (AHC) between 1/1/2016 and 12/31/2017. MAIN MEASURES: Sensitivity, specificity, positive predictive value, C-statistic, and risk threshold score of the model. KEY RESULTS: The final model was strongly discriminative (C-statistic = 0.90) and had a good fit (Hosmer-Lemeshow goodness-of-fit test [X 2 (8) =6.26, p = 0.62]). The positive predictive value for 30-day in-hospital death was 68%; AUC-ROC was 0.90 (95% confidence interval 0.89-0.92, p < 0.0001). We identified a risk threshold score of −2.19 that had a maximum sensitivity (79.87%) and specificity (85.24%) in the derivation and validation sample (sensitivity: 75.00%, specificity: 85.71%). In the validation sample, 34.40% (354/1029) of the patients above this threshold died compared to only 2.83% (118/4166) deaths below this threshold. CONCLUSION: This model can use EMR and administrative data within 24 h of transfer to predict the risk of 30day in-hospital mortality with reasonable accuracy among seriously ill transferred patients.

The Indiana Learning Health System Initiative: Early experience developing a collaborative, regional learning health system

“…In the early mortality project (Table 3), for instance, ILHSI investigators partnered with the ASPIRE program to conduct a project that addressed an immediate priority for IUH clinical leadership. 30 Fourth, our experience has confirmed our initial expectation that the LHS must be data-driven, a strength of the RI and an important aspect of the collaborative nature of the ILHSI. A major challenge facing collaborative LHSs is interoperability of data.…”

“…As one example, we partnered with the ASPIRE training program to mentor a junior hospitalist to conduct a project developing a model for predicting early in‐hospital mortality among patients transferred in to the IU Academic Health Center. 30 This project is being carried forward in a current implementation project to identify patients for whom early palliative care consultation is indicated, and to determine barriers and facilitators to the use of this model among hospitalists.…”

“…In the early mortality project (Table 3 ), for instance, ILHSI investigators partnered with the ASPIRE program to conduct a project that addressed an immediate priority for IUH clinical leadership. 30 …”

Using Electronic Health Record Mortality Data to Promote Goals-of-Care Discussions in Seriously Ill Transferred Patients: A Pilot Study

“…A complete description of the model's development and evaluation are published elsewhere. 24 In this study, we hypothesized that an intervention involving real-time communication of our model's all-cause 30-day inpatient mortality risk with the primary hospitalists could promote early GOCD in seriously ill transferred patients. 24…”

“…Patients were eligible for inclusion if they were general medical patients, were 18 years of age or older, transferred from outside facilities, admitted to the hospitalist service, had decision-making capacity or an assigned health care representative/surrogate, and were identified by our model to be at risk for 30-day inpatient mortality. 24 Recruitment was limited to patients admitted between Saturday 1:00 a.m. and Friday 8:00 a.m. during the intervention period to align with our measurement of whether palliative care consultations occurred within 72 hours of admission as there are no palliative care consultations at the hospital over the weekend. Patients were excluded if they were admitted to locum hospitalist teams, if the primary hospitalist teams were primary researcher for this study or if they were involved in concurrent research studies (i.e., Red and Orange teams), if the patients had been transferred directly to our ICU on admission, if the patients were admitted from our ED, if the patients died within 24 hours of transfer, and if they had a documented GOCD or palliative care consultation at an outside hospital before transfer without any changes to the plan of care.…”