Our Blog

Abstract: Hospice Analytics Technology Helps Reduce Early Death On Hospice By 30%

Posted by Michael Faron on March 18, 2015

The Impact of Medalogix Bridge on Early Mortality for Hospice Patients

Michael Faron, Medalogix LLC, Nashville, TN

Reviewed by Christine Lai, Ph.D., Vanderbilt University[1]



Patients referred to hospice for appropriate medical care are typically terminally ill. The national mean and median of the length of service of annual hospice care discharges in 2007 were 65 days and 16 days respectively as reported in the National Health Statistics Reports[2]. The most common discharge disposition was death (84.3%)[3]. Advance care planning is important to the patients, their family, clinical professionals and administrative staff at hospice agencies.

Medalogix Bridge is designed to be used as a clinical decision support tool to assist clinicians in transitioning the right Home Health patients to hospice at the right time. Physicians, nurses and home health administrators use predictive model risk scores of patients as an input for deciding whether to put the patient into the Medalogix Bridge Work Flow (WF) to start a process of monitoring, evaluating and final referral to hospice transfer.

 The purpose of the study is to test the hypothesis that the adoption of the Medalogix Bridge for referring patients to hospice care can reduce the early death rate (death within 7 days) on hospice. If data supports the hypothesis, the Bridge WF value adds to the process of hospice referral allowing more time for the imperative advance care planning.

Design and Methodology

Patients of a large hospice agency in the southeast United States discharged between April 14, 2014 and February 13, 2015 are the subjects of the study. There are many sources from where patients are referred to the hospice. The study only considers those patients sourced from the home health agency and hospice of the same company, that is, only internal transfers are considered in the study mainly due to data availability. There are a total of 1,729 hospice patients among which a total of 652 were referred to the hospice after receiving care under Medalogix Bridge (as the treatment group) while on Home Health. The other 1,077 were not under any Medalogix Bridge monitoring or referral (as the control group).[4] Observational data is sourced from the hospice agency on patient’s demographics, functionality and physical condition severity.

Any death within 7 days of the hospice admission is labeled as early death (the outcome measure of the study). Patients who deceased after the 7th day or survived all the way when discharged from the hospice are considered not having the study outcome event. In order to test the hypothesis in a similar setting as randomized controlled trials (RCTs), a propensity score one-to-one matching methodology is applied. Using a logistic regression, the patient's profile in the treatment group is identified and then matched with those of the control group with a similar propensity score.[5] After matching, the hypothesis testing then provides a valid estimate of treatment effect because the research method compares patients with similar observed characteristics, all of whom are potential candidates for the treatment.


There are 13 covariates in the propensity models. At 10% level of significance, all the covariates in the model except ‘the Weekly Indicator of Incident’ are not significant, suggesting that the treatment and control groups are similar in terms of the covariates in the model. Besides age and gender, all the non-significant covariates are measures on patient's functionality and frailty. As both treatment and control group patients are under hospice care, they are unsurprisingly homogenous in terms of these health status covariates. The similarity of the treatment and control groups prior to the matching suggests that there was limited selection bias in the hospice study sample.


The early death rate of the treatment group was 11.7% (76 out of 652 patients) while it was 15.8% (170 out of 1,077 patients) in the control group prior to the matching. Seemingly, patients referred via Medalogix Bridge had a lower early death rate than those via elsewhere, suggesting the Bridge patients were 26% less likely than their counterparts to expire within 7 days after admission to hospice. The odds ratio is of greater magnitude after matching: Bridge patients (early death rate 11.5% from 74 out of 642 patients) were 30% less likely to have early death than those non-Bridge patients (early death rate 16.5% from 106 out of 642). The difference in early death rates is significant before and after matching at 5% and 1% level of significance respectively[6].


The hypothesis that patients referred to hospice after receiving care under Medalogix Bridge during their Home Health stay have a lower early death rate (within 7 days after admission) than those patients not receiving care under Bridge is supported by the data. Propensity score one-to-one matching is conducted to remove possible selection bias. The adoption of Medalogix Bridge for hospice referral can reduce early death as much as 30%. Patients also can have appropriate and early hospice care allowing more available time for patients, their family, as well as the administrative staff to have proper advance care planning.


  1. Christine Caffrey, Ph.D. et al. “Home Health Care and Discharged Hospice Care Patients: United States, 2000 and 2007,” National Health Statistics Reports, Centers for Disease Control and Prevention, Number 38, April 2011.

[1] Disclaimer: The contents of the abstract are of the author’s sole responsibility. They do not represent the views of Vanderbilt University Medical Center. Any inquiries related to the abstract and all permission to cite should be directed to the author.

[2] Christine Caffrey, Ph.D. et al. “Home Health Care and Discharged Hospice Care Patients: United States, 2000 and 2007,” National Health Statistics Reports, Centers for Disease Control and Prevention, Number 38, April 2011.

[3] The remaining 15.6% were discharged either back home, to another hospice, to an inpatient care, etc. If the number of deceased at elsewhere (within a short time) is included, the death rate could be higher.

[4] Due to the small sample size, we went back one calendar year before the inception of WF but in the same monthly time frame. It then added 805 more cases to the group without WF.

[5] The threshold difference is 0.2 of a standard deviation of the logit scores.

[6] The significance level is per Chi-square test and McNemar’s test before and after matching respectively.

Topics: Data and Analytics Review

How post acute providers can change the way we die

Subscribe to Email Updates

Recent Posts