Machine learning functional impairment classification with electronic health record data.

BACKGROUND: Poor functional status is a key marker of morbidity, yet is not routinely captured in clinical encounters. We developed and evaluated the accuracy of a machine learning algorithm that leveraged electronic health record (EHR) data to provide a scalable process for identification of functional impairment. METHODS: We identified a cohort of patients with an electronically captured screening measure of functional status (Older Americans Resources and Services ADL/IADL) between 2018 and 2020 (N = 6484). Patients were classified using unsupervised learning K means and t-distributed Stochastic Neighbor Embedding into normal function (NF), mild to moderate functional impairment (MFI), and severe functional impairment (SFI) states. Using 11 EHR clinical variable domains (832 variable input features), we trained an Extreme Gradient Boosting supervised machine learning algorithm to distinguish functional status states, and measured prediction accuracies. Data were randomly split into training (80%) and test (20%) sets. The SHapley Additive Explanations (SHAP) feature importance analysis was used to list the EHR features in rank order of their contribution to the outcome. RESULTS: Median age was 75.3 years, 62% female, 60% White. Patients were classified as 53% NF (n = 3453), 30% MFI (n = 1947), and 17% SFI (n = 1084). Summary of model performance for identifying functional status state (NF, MFI, SFI) was AUROC (area under the receiving operating characteristic curve) 0.92, 0.89, and 0.87, respectively. Age, falls, hospitalization, home health use, labs (e.g., albumin), comorbidities (e.g., dementia, heart failure, chronic kidney disease, chronic pain), and social determinants of health (e.g., alcohol use) were highly ranked features in predicting functional status states. CONCLUSION: A machine learning algorithm run on EHR clinical data has potential utility for differentiating functional status in the clinical setting. Through further validation and refinement, such algorithms can complement traditional screening methods and result in a population-based strategy for identifying patients with poor functional status who need additional health resources.

Assessment of Clinical Palliative Care Trigger Status vs Actual Needs Among Critically Ill Patients and Their Family Members.

Importance: Palliative care consultations in intensive care units (ICUs) are increasingly prompted by clinical characteristics associated with mortality or resource utilization. However, it is not known whether these triggers reflect actual palliative care needs. Objective: To compare unmet needs by clinical palliative care trigger status (present vs absent). Design, Setting, and Participants: This prospective cohort study was conducted in 6 adult medical and surgical ICUs in academic and community hospitals in North Carolina between January 2019 and September 2020. Participants were consecutive patients receiving mechanical ventilation and their family members. Exposure: Presence of any of 9 common clinical palliative care triggers. Main Outcomes and Measures: The primary outcome was the Needs at the End-of-Life Screening Tool (NEST) score (range, 0-130, with higher scores reflecting greater need), which was completed after 3 days of ICU care. Trigger status performance in identifying serious need (NEST score ≥30) was assessed using sensitivity, specificity, positive and negative likelihood ratios, and C statistics. Results: Surveys were completed by 257 of 360 family members of patients (71.4% of the potentially eligible patient-family member dyads approached) with a median age of 54.0 years (IQR, 44-62 years); 197 family members (76.7%) were female, and 83 (32.3%) were Black. The median age of patients was 58.0 years (IQR, 46-68 years); 126 patients (49.0%) were female, and 88 (33.5%) were Black. There was no difference in median NEST score between participants with a trigger present (45%) and those with a trigger absent (55%) (21.0; IQR, 12.0-37.0 vs 22.5; IQR, 12.0-39.0; P = .52). Trigger presence was associated with poor sensitivity (45%; 95% CI, 34%-55%), specificity (55%; 95% CI, 48%-63%), positive likelihood ratio (1.0; 95% CI, 0.7-1.3), negative likelihood ratio (1.0; 95% CI, 0.8-1.2), and C statistic (0.50; 95% CI, 0.44-0.57). Conclusions and Relevance: In this cohort study, clinical palliative care trigger status was not associated with palliative care needs and no better than chance at identifying the most serious needs, which raises questions about an increasingly common clinical practice. Focusing care delivery on directly measured needs may represent a more person-centered alternative.

Utilization of Hospice Services in a Population of Patients With Huntington's Disease.

CONTEXT: Although the early and middle stages of Huntington's disease (HD) and its complications have been well described, less is known about the course of late-stage illness. In particular, little is known about the population of patients who enroll in hospice. OBJECTIVES: Our goal is to describe the characteristics of patients with HD who enrolled in hospice. METHODS: This is a retrospective cohort study of electronic medical record data from 12 not-for-profit hospices in the United States from 2008 to 2012. RESULTS: Of the 164,032 patients admitted to these hospices, 101 (0.06%) had a primary diagnosis of HD. Their median age was 57 (IQR 48-65) and 53 (52.5%) were women. Most patients were cared for by a spouse (n = 36, 36.6%) or adult child (n = 20, 19.8%). At the time of admission, most patients were living either at home (n = 39, 38.6%) or in a nursing home (n = 41, 40.6%). All were either bedbound or could ambulate only with assistance. The most common symptom reported during enrollment in hospice was pain (n = 34, 33.7%) followed by anxiety (n = 30, 29.7%), nausea (n = 18, 17.8%), and dyspnea (n = 10, 9.9%). Patients had a median length of stay in hospice of 42 days, which was significantly longer than that of other hospice patients in the sample (17 days), P < 0.001. Of the 101 patients who were admitted to hospice, 73 died, 11 were still enrolled at the time of data analysis, and 17 left hospice either because they no longer met eligibility criteria (n = 14, 13.7%) or because they decided to seek treatment for other medical conditions (n = 3, 3.0%). Of the 73 patients who died while on hospice, most died either in a nursing home (n = 29; 40%) or a hospital (n = 27; 37%). Seventeen patients (23%) died at home. No patient that started in a facility died at home. CONCLUSION: Patients with HD are admitted to hospice at a younger age compared with other patients (57 vs. 76 years old) but have a significant symptom burden and limited functional status. Although hospice care emphasizes the importance of helping patients to remain in their homes, only a minority of these patients were able to die at home.

Palliative Care Planner: A Pilot Study to Evaluate Acceptability and Usability of an Electronic Health Records System-integrated, Needs-targeted App Platform.

RATIONALE: The quality and patient-centeredness of intensive care unit (ICU)-based palliative care delivery is highly variable. OBJECTIVE: To develop and pilot an app platform for clinicians and ICU patients and their family members that enhances the delivery of needs-targeted palliative care. METHODS: In the development phase of the study, we developed an electronic health record (EHR) system-integrated mobile web app system prototype, PCplanner (Palliative Care Planner). PCplanner screens the EHR for ICU patients meeting any of five prompts (triggers) for palliative care consultation, allows families to report their unmet palliative care needs, and alerts clinicians to these needs. The evaluation phase included a prospective before/after study conducted at a large academic medical center. Two control populations were enrolled in the before period to serve as context for the intervention. First, 25 ICU patients who received palliative care consults served as patient-level controls. Second, 49 family members of ICU patients who received mechanical ventilation for at least 48 hours served as family-level controls. Afterward, 14 patients, 18 family members, and 10 clinicians participated in the intervention evaluation period. Family member outcomes measured at baseline and 4 days later included acceptability (Client Satisfaction Questionnaire [CSQ]), usability (Systems Usability Scale [SUS]), and palliative care needs, assessed with the adapted needs of social nature, existential concerns, symptoms, and therapeutic interaction (NEST) scale; the Patient-Centeredness of Care Scale (PCCS); and the Perceived Stress Scale (PSS). Patient outcomes included frequency of goal concordant treatment, hospital length of stay, and discharge disposition. RESULTS: Family members reported high PCplanner acceptability (mean CSQ, 14.1 [SD, 1.4]) and usability (mean SUS, 21.1 [SD, 1.7]). PCplanner family member recipients experienced a 12.7-unit reduction in NEST score compared with a 3.4-unit increase among controls (P = 0.002), as well as improved mean scores on the PCCS (6.6 [SD, 5.8]) and the PSS (-0.8 [SD, 1.9]). The frequency of goal-concordant treatment increased over the course of the intervention (n = 14 [SD, 79%] vs. n = 18 [SD, 100%]). Compared with palliative care controls, intervention patients received palliative care consultation sooner (3.9 [SD, 2.7] vs. 6.9 [SD, 7.1] mean days), had a shorter mean hospital length of stay (20.5 [SD, 9.1] vs. 22.3 [SD, 16.0] patient number), and received hospice care more frequently (5 [36%] vs. 5 [20%]), although these differences were not statistically significant. CONCLUSIONS: PCplanner represents an acceptable, usable, and clinically promising systems-based approach to delivering EHR-triggered, needs-targeted ICU-based palliative care within a standard clinical workflow. A clinical trial in a larger population is needed to evaluate its efficacy.

Rationale and Design of the Randomized Evaluation of Default Access to Palliative Services (REDAPS) Trial.

The substantial nationwide investment in inpatient palliative care services stems from their great promise to improve patient-centered outcomes and reduce costs. However, robust experimental evidence of these benefits is lacking. The Randomized Evaluation of Default Access to Palliative Services (REDAPS) study is a pragmatic, stepped-wedge, cluster randomized trial designed to test the efficacy and costs of specialized palliative care consultative services for hospitalized patients with advanced chronic obstructive pulmonary disease, dementia, or end-stage renal disease, as well as the overall effectiveness of ordering such services by default. Additional aims are to identify the types of services that are most beneficial and the types of patients most likely to benefit, including comparisons between ward and intensive care unit patients. We hypothesize that patient-centered outcomes can be improved without increasing costs by simply changing the default option for palliative care consultation from opt-in to opt-out for patients with life-limiting illnesses. Patients aged 65 years or older are enrolled at 11 hospitals using an integrated electronic health record. As a pragmatic trial designed to enroll between 12,000 and 15,000 patients, eligibility is determined using a validated, electronic health record-based algorithm, and all outcomes are captured via the electronic health record and billing systems data. The time at which each hospital transitions from control, opt-in palliative care consultation to intervention, opt-out consultation is randomly assigned. The primary outcome is a composite measure of in-hospital mortality and length of stay. Secondary outcomes include palliative care process measures and clinical and economic outcomes. Clinical trial registered with www.clinicaltrials.gov (NCT02505035).

Differences in Terminal Hospitalization Care Between U.S. Men and Women.

CONTEXT: In many settings, men and women receive different care. OBJECTIVES: We sought to determine whether men and women receive different care during terminal hospitalizations. METHODS: We analyzed data of 98,314 adult patients who died while hospitalized in 458 acute care hospitals in the U.S. during 2011. We examined sex-based differences in lengths of stay (LOS), resuscitation status, and intensive interventions and processes of care, adjusting for patient- and hospital-level characteristics. RESULTS: Women represented half of the sample (48,509; 49.34%), were older than men (73.8 vs. 70.6 years, P < 0.0001), and less likely to be married (27.7% vs. 48.3%, P < 0.001). Among all patients, median LOS was four days (interquartile range 2-10); 19.1% of subjects received cardiopulmonary resuscitation; 37.6% had a do-not-resuscitate order during the admission; and 51.6% received mechanical ventilation. Compared with men, women had slightly shorter hospitalizations (adjusted LOS: -0.16 days; 95% CI -0.19, -0.12) and were more likely to have a do-not-resuscitate order (odds ratio [OR] 1.08; 95% CI 1.05, 1.11). Women remained less likely to receive care in an intensive care unit (OR 0.95; 95% CI 0.93, 0.98), cardiopulmonary resuscitation (OR 0.83; 95% CI 0.80, 0.86), mechanical ventilation (OR 0.94; 95% CI 0.91, 0.97), hemodialysis (adjusted OR 0.81; 95% CI 0.78, 0.86), or surgical procedures (OR 0.88; 95% CI 0.84, 0.93). CONCLUSION: Men who die in hospitals receive more aggressive care than women. Further research should examine potential causes of this overall pattern.

Development of an Assessment to Examine Training of the Hospice Primary Caregiver.

BACKGROUND: Key to high-quality care of dying hospice patients at home is whether the hospice provides adequate training so the caregiver can safely care for the patient. OBJECTIVE: The study objective was to develop and validate a survey of hospice training for caregivers to ensure safe, high-quality care in the home setting. DESIGN: Our survey design was cross-sectional. Bereaved respondents of individuals who died at home under the care of hospice were surveyed three to six months postdeath. MEASUREMENTS: Items were developed based on advice of an expert panel, focus groups of hospice caregivers, and literature review, with 12 items developed for testing and examining 8 key processes of care. We examined the validity and reliability of the assessment using factor analysis, correlational analyses, and multivariable modeling. RESULTS: Our sample consisted of 262 primary caregivers (mean age 62.4, 76.7% female, 58.8% non-Hispanic white). Six questions focused on providing the caregiver with information, while another six focused on the training that hospice provided. Based on model fit and Cronbach's alpha, we dropped the information items. The items that examined hospice training demonstrated a one-factor solution and a Cronbach's of 0.90. We examined correlations of the multi-item composite with overall rating of quality of care (0.53), overall distress (0.31), and whether the respondent would recommend this hospice to others (0.49). There were no significant sociodemographic correlates of concerns with training. CONCLUSIONS: Sufficient preliminary reliability and validity warrants further testing of this composite to examine the adequacy of training provided to family members to care safely for the patient.

The Impact of Inpatient Palliative Care Consultations on 30-Day Hospital Readmissions.

BACKGROUND: Inpatient palliative care consultations have been shown to reduce acute care utilization by reducing length of stay, but less is known about their impact on subsequent costs including hospital readmissions. OBJECTIVE: The study's objective was to examine the impact of inpatient palliative care consultations on 30-day hospital readmissions to a large urban academic medical center. METHODS: The hospital's electronic medical record system was used to identify all live discharges between August 2013 and November 2014. After adjusting for a propensity score, readmission rates were compared between palliative care and usual care groups. RESULTS: Of the 34,541 hospitalizations included in the study, 1430 (4.1%) involved a palliative care consult. After adjusting for the propensity score, patients seen by palliative care had a lower 30-day readmission rate-adjusted odds ratio (AOR) 0.66, 0.55-0.78; p<0.001. Adjusted rates were 10.3% (95% confidence interval [CI] 8.9%-12.0%) for palliative care and 15.0% (95% CI 14.4%-15.4%) for usual care. Among all palliative care patients, consultations that involved goals of care discussions were associated with a lower readmission rate (AOR 0.36, 0.27-0.48; p<0.001), but consultations involving symptom management were not (AOR 1.05, 0.82-1.35; p=0.684). CONCLUSIONS: Palliative care palliative care consultations facilitate goals discussions, which in turn are associated with reduced rates of 30-day readmissions.

Comparison of pediatric and adult hospice patients using electronic medical record data from nine hospices in the United States, 2008-2012.

BACKGROUND: Most pediatric hospice patients receive services from agencies typically oriented to adults. Information regarding how pediatric hospice patients differ from adult hospice patients is lacking. OBJECTIVE: We aim to assess differences between pediatric and adult hospice patients regarding patient characteristics and outcomes. METHODS: We compiled a retrospective inception cohort of patients enrolled at nine hospices in the CHOICE network (Coalition of Hospices Organized to Investigate Comparative Effectiveness) between August 1, 2008 and June 30, 2012. Measurements included patient characteristics and outcomes, including discharge from hospice and site of death. RESULTS: Among 126,620 hospice patients, 986 (0.8%) were 18 years of age or younger. Pediatric patients were less likely to have an admitting diagnosis of cancer (odds ratio [OR] 0.62; 95% confidence interval [CI]: 0.54-0.72). Although children were less likely to use oxygen at enrollment (OR 0.31; 95% CI: 0.26-0.37), they were more likely to have an enteral feeding tube (OR 4.04; 95% CI: 3.49-4.67). Pediatric patients were half as likely as adults to have a do-not-resuscitate order (DNR) order upon hospice enrollment (OR 0.52; 95% CI: 0.46-0.59). The average hospice length of stay for pediatric patients was longer than that of adults (103 days versus 66 days, p<0.001). Children were more likely to leave hospice care (OR 2.59; 95% CI: 2.00-3.34), but among patients who died while enrolled in hospice, pediatric patients were more likely to die at home (OR 3.25; 95% CI: 2.27-3.88). CONCLUSIONS: Pediatric hospice patients differ from adult patients in their broader range of underlying diagnoses and their use of hospice services.

The quality imperative for palliative care.

Palliative medicine must prioritize the routine assessment of the quality of clinical care we provide. This includes regular assessment, analysis, and reporting of data on quality. Assessment of quality informs opportunities for improvement and demonstrates to our peers and ourselves the value of our efforts. In fact, continuous messaging of the value of palliative care services is needed to sustain our discipline; this requires regularly evaluating the quality of our care. As the reimbursement mechanisms for health care in the U.S. shift from fee-for-service to fee-for-value models, palliative care will be expected to report robust data on quality of care. We must move beyond demonstrating to our constituents (including patients and referrers), "here is what we do," and increase the focus on "this is how well we do it" and "let us see how we can do it better." It is incumbent on palliative care professionals to lead these efforts. This involves developing standardized methods to collect data without adding additional burden, comparing and sharing our experiences to promote discipline-wide quality assessment and improvement initiatives, and demonstrating our intentions for quality improvement on the clinical frontline.

Hospice admissions for cancer in the final days of life: independent predictors and implications for quality measures.

PURPOSE: To define patient characteristics associated with hospice enrollment in the last 3 days of life, and to describe adjusted proportions of patients with late referrals among patient subgroups that could be considered patient-mix adjustment variables for this quality measure. METHODS: Electronic health record-based retrospective cohort study of patients with cancer admitted to 12 hospices in the Coalition of Hospices Organized to Investigate Comparative Effectiveness network. RESULTS: Of 64,264 patients admitted to hospice with cancer, 10,460 (16.3%) had a length of stay ≤ 3 days. There was significant variation among hospices (range, 11.4% to 24.5%). In multivariable analysis, among patients referred to hospice, patients who were admitted in the last 3 days of life were more likely to have a hematologic malignancy, were more likely to be male and married, and were younger (age < 65 years). Patients with Medicaid or self-insurance were less likely to be admitted to hospice within 3 days of death. CONCLUSION: Quality measures of hospice lengths of stay should include patient-mix adjustments for type of cancer and site of care. Patients with hematologic malignancies are at especially increased risk for late admission to hospice.

Family perceptions of quality of hospice care in the nursing home.

CONTEXT: Nursing homes (NHs) are increasingly the site of hospice care. High quality of care is dependent on successful NH-hospice collaboration. OBJECTIVES: To examine bereaved family members' perceptions of NH-hospice collaborations in terms of what they believe went well or could have been improved. METHODS: Focus groups were conducted with bereaved family members from five diverse geographic regions, and included participants from inner city and rural settings, with oversampling of African Americans. RESULTS: A total of 28 participants (14.8% African American, mean age 61.4 years) identified three major aspects of collaboration as important to care delivery. First, most (67.9%) voiced concerns with knowing who (NH or hospice) is responsible for which aspects of patient care. Second, nearly half (42.9%) stated concern about information coordination between the NH and hospice. Finally, 67.9% of the participants mentioned the need for hospice to advocate for high-quality care rather than their having to directly do so on behalf of their family members. CONCLUSION: The important concerns raised by bereaved family members about NH-hospice collaboration have been incorporated into the revised Family Evaluation of Hospice Care, a post-death survey used to evaluate quality of hospice care.

The "comfortable dying" measure: how patient characteristics affect hospice pain management quality scores.

BACKGROUND: All hospices were required by the Centers for Medicare and Medicaid Services (CMS) to collect the "Comfortable Dying" measure in 2012 (National Quality Forum measure #0209). However, it is not known how scores on this measure are affected by patient characteristics. It is important to identify these characteristics so that a hospice's case mix can be taken into account when interpreting its scores. OBJECTIVE: Our aim was to describe the implementation of the NQF #0209 measure in 10 hospices and to identify patient characteristics associated with scores. METHODS: We conducted an electronic health record (EHR)-based retrospective cohort study of patients in 10 hospices in the United States. The main outcome measure was the proportion of patients with pain that made them uncomfortable whose pain was controlled within 48 hours. RESULTS: A total of 4157 patients were eligible for an initial pain assessment. Of those who reported pain (n=1992), 1152 (58%) reported having their pain controlled on the follow-up assessment. In a multivariable regression model, clustered by hospice, six variables were independently associated with pain control. These included age (adjusted odds ratio [OR] 1.02; 95% confidence interval [CI] 1.02-1.03, p=0.003), a cancer diagnosis (OR 1.37; 95% CI 1.20-1.53, p=0.008), initial care in an inpatient unit (OR 1.28; 95% CI 1.08-1.47, p=0.031), presence of a Foley catheter (OR 1.40; 95% CI 1.15-1.59, p=0.038), use of opioid medication (OR 1.34; 95% CI 1.03-1.74, p=0.027), and higher Palliative Performance Scale (PPS) score (OR 1.02; 95% CI 1.01-1.03, p<0.001). Presence of a Stage 2 pressure ulcer was independently associated with worse pain control (OR 0.63; 95% CI 0.31-0.96, p=0.012). CONCLUSIONS: Several patient characteristics are associated with #0209 pain scores. As hospices are increasingly required to report quality measures, it will be essential to understand how their scores are affected by case mix.

All hospice patients are not equal: development of a visit-based acuity index.

BACKGROUND: Although hospices need to be able to anticipate patient acuity, there are currently no published models that predict the frequency of visits that a new hospice patient is likely to receive. OBJECTIVES: To identify patient characteristics that are associated with the frequency of health care provider visits in the first 8 days of hospice care. METHODS: An electronic health record (EHR)-based retrospective cohort study was conducted in seven hospice programs in the United States. Participants were 35,232 patients who were admitted between October 1, 2008 and May 31, 2011 and received hospice care at home on the day of enrollment. The main outcome measure was the average number of visits per day by nurses, social workers, chaplains, and home health aides in the first 8 days of home hospice care (day of admission plus up to 7 subsequent days). RESULTS: In a mixed effects regression model, 14 independent predictors of visit frequency were identified. For instance, several demographic characteristics were associated with more frequent visits, as were lower Palliative Performance Scale (PPS) scores (<40: 1.78 visits/day, 95% confidence interval [CI] 1.74-1.82 versus 40-60: 1.65 visits/day, 95% CI 1.61-1.69 versus >60: 1.41 visits/day, 95% CI 1.36-1.47; p<0.001), the presence of pain (pain: 1.77 visits/day; 95% CI 1.72-1.82 versus no pain: 1.44 visits/day, 95% CI .39-1.59; p<0.001). Patients admitted to home hospice from a hospital also received more frequent visits compared with other patients (hospital: 1.73 visits/day, 95% CI 1.67-1.79 versus home: 1.42 visits/day, 95% CI 1.40-1.44; p<0.001). CONCLUSIONS: An acuity index based on these variables could help hospices to better anticipate patient needs and staff workload, and could be used to guide strategic planning as hospices take part in accountable care organizations (ACOs).

Effect of the Medicare face-to-face visit requirement on hospice utilization.

BACKGROUND: Although many patients enter hospice close to death, some enroll for more than six months. In 2011 the U.S. Centers for Medicare and Medicaid Services (CMS) required that these long-stay patients receive a face-to-face visit by a physician or nurse practitioner to ensure that they continue to meet eligibility criteria. OBJECTIVES: This study proposed to determine whether the face-to-face visit requirement increased the rate at which patients were decertified from hospice. DESIGN: The study was a retrospective cohort study in six U.S. hospices. Decertification from hospice within 10 months of enrollment was measured. RESULTS: Of 23,638 patients, 11,788 (49.9%) would have been affected by the face-to-face requirement. In bivariate analysis, there was a significant decrease in the decertification rate after the requirement was implemented-371/11,788 (3.2%) versus 578/11,850 (4.9%); odds ration (OR): 0.63; 95% CI 0.55-0.72; p<0.001. In a multivariable logistic regression model adjusting for changes in patient characteristics and clustered by hospice, there was still a reduction in decertifications-3.4% versus 5.2%; OR 0.67; 95% CI 0.47-0.97; p=0.034. Although the impact of the face-to-face requirement varied among hospices, all hospices had a decrease in decertification rates (absolute adjusted reduction between 1.4% and 3.6%). CONCLUSIONS: The face-to-face requirement may decrease hospice discharges, contrary to its intention.