PURPOSE: Monitoring and improving the quality of palliative and end-of-life cancer care remain pressing needs in the United States. Among existing measures that assess the quality of palliative and end-of-life care, many operationalize similar concepts. We identified existing palliative care process measures and synthesized these measures to aid stakeholder prioritization that will facilitate health system implementation in patients with advanced cancer.
METHODS: We reviewed MEDLINE/PubMed-indexed articles for process quality measures related to palliative and end-of-life care for patients with advanced cancer, supplemented by expert input. Measures were inductively grouped into "measure concepts" and higher-level groups.
RESULTS: Literature review identified 226 unique measures from 23 measure sources, which we grouped into 64 measure concepts within 12 groups. Groups were advance care planning (11 measure concepts), pain (7), dyspnea (9), palliative care-specific issues (6), other specific symptoms (17), comprehensive assessment (2), symptom assessment (1), hospice/palliative care referral (1), spiritual care (2), mental health (5), information provision (2), and culturally appropriate care (1).
CONCLUSION: Measure concepts covered the spectrum of care from acute symptom management to advance care planning and psychosocial needs, with variability in the number of measure concepts per group. This taxonomy of process quality measure concepts can be used by health systems seeking stakeholder input to prioritize targets for improving palliative and end-of-life care quality in patients with advanced cancer.
INTRODUCTION: Advance care planning (ACP) is associated with improved health outcomes for patients with cancer, and its absence is associated with unfavourable outcomes for patients and their caregivers. However, older adults do not complete ACP at expected rates due to patient and clinician barriers. We present the original design, methods and rationale for a trial aimed at improving ACP for older patients with advanced cancer and the modified protocol in response to changes brought by the COVID-19 pandemic.
METHODS AND ANALYSIS: The Advance Care Planning: Promoting Effective and Aligned Communication in the Elderly study is a pragmatic, stepped-wedge cluster randomised trial examining a Comprehensive ACP Program. The programme combines two complementary evidence-based interventions: clinician communication skills training (VitalTalk) and patient video decision aids (ACP Decisions). We will implement the programme at 36 oncology clinics across three unique US health systems. Our primary outcome is the proportion of eligible patients with ACP documentation completed in the electronic health record. Our secondary outcomes include resuscitation preferences, palliative care consultations, death, hospice use and final cancer-directed therapy. From a subset of our patient population, we will collect surveys and video-based declarations of goals and preferences. We estimate 11 000 patients from the three sites will be enrolled in the study.
ETHICS AND DISSEMINATION: Regulatory and ethical aspects of this trial include Institutional Review Board (IRB) approval via single IRB of record mechanism at Dana-Farber Cancer Institute, Data Use Agreements among partners and a Data Safety and Monitoring Board. We plan to present findings at national meetings and publish the results.
Background: Serious illness conversations are part of advance care planning (ACP) and focus on prognosis, values, and goals in patients who are seriously ill. To be maximally effective, such conversations must be documented accurately and be easily accessible.
Objectives: The two coprimary objectives of the study were to assess concordance between written documentation and recorded audiotaped conversations, and to evaluate adherence to the Serious Illness Conversation Guide questions.
Methods: Data were obtained as part of a trial in patients with advanced cancer. Clinicians were trained to use a guide to conduct and document serious illness conversations. Conversations were audiotaped. Two researchers independently compared audiorecordings with the corresponding documentation in an electronic health record (EHR) template and free-text progress notes, and rated the degree of concordance and adherence.
Results: We reviewed a total of 25 audiorecordings. Clinicians addressed 87% of the conversation guide elements. Prognosis was discussed least frequently, only in 55% of the patients who wanted that information. Documentation was fully concordant with the conversation 43% of the time. Concordance was best when documenting family matters and goals, and least frequently concordant when documenting prognostic communication. Most conversations (64%) were documented in the template, a minority (28%) only in progress notes and two conversations (8%) were not documented. Concordance was better when the template was used (62% vs. 28%).
Conclusion: Clinicians adhered well to the conversation guide. However, key information elicited was documented and fully concordant less than half the time. Greater concordance was observed when clinicians used a prespecified template. The combined use of a guide and EHR template holds promise for ACP conversations.
CONTEXT: The Trauma Quality Improvement Program Best Practice Guidelines recommend palliative care (PC) concurrent with restorative treatment for patients with life-threatening injuries. Measuring PC delivery is challenging: administrative data are nonspecific, and manual review is time intensive.
OBJECTIVES: To identify PC delivery to patients with life-threatening trauma and compare the performance of natural language processing (NLP), a form of computer-assisted data abstraction, to administrative coding and gold standard manual review.
METHODS: Patients 18 years and older admitted with life-threatening trauma were identified from two Level I trauma centers (July 2016-June 2017). Four PC process measures were examined during the trauma admission: code status clarification, goals-of-care discussion, PC consult, and hospice assessment. The performance of NLP and administrative coding were compared with manual review. Multivariable regression was used to determine patient and admission factors associated with PC delivery.
RESULTS: There were 76,791 notes associated with 2093 admissions. NLP identified PC delivery in 33% of admissions compared with 8% using administrative coding. Using NLP, code status clarification was most commonly documented (27%), followed by goals-of-care discussion (18%), PC consult (4%), and hospice assessment (4%). Compared with manual review, NLP performed more than 50 times faster and had a sensitivity of 93%, a specificity of 96%, and an accuracy of 95%. Administrative coding had a sensitivity of 21%, a specificity of 92%, and an accuracy of 68%. Factors associated with PC delivery included older age, increased comorbidities, and longer intensive care unit stay.
CONCLUSION: NLP performs with similar accuracy with manual review but with improved efficiency. NLP has the potential to accurately identify PC delivery and benchmark performance of best practice guidelines.
Background: Periprocedural providers are encountering more patients with code status limitations (CSLs) regarding their preferences for resuscitation and life-sustaining treatment who choose to undergo palliative procedures. Surgical and anesthesia guidelines for preprocedural reconsideration of CSLs have been available for several years, but it is not known whether they are being followed in practice.
Objective: We assessed compliance with existing guidelines for patients undergoing venting gastrostomy tube (VGT) for malignant bowel obstruction (MBO), serving as an example of a palliative procedure received by patients near the end of life.
Design: Code status was determined at admission and throughout the hospitalization by chart review. Documentation of code status discussions (CSDs) was identified from provider notes and compared with existing guidelines.
Setting/subjects: An institutional database retrospectively identified patients who underwent VGT placement for MBO at two academic hospitals (2014-2015).
Measurements: We identified 53 patients who underwent VGT placement for MBO. Interventional radiologists performed 88% of these procedures. Other periprocedural providers involved in these cases included surgeons, gastroenterologists, anesthesiologists, and sedation nurses.
Results: CSLs were documented before the procedure in only 43% of cases, and a documented CSD with a periprocedural provider was identified in only 22% of CSL cases. Of all VGT placements performed in patients with CSLs before the procedure, only 13% were compliant with the guidelines of preprocedural reconsideration of CSLs.
Conclusions: Increased compliance with guidelines published by the American Society of Anesthesiologists, the American College of Surgeons, and the Association of Perioperative Registered Nurses is necessary to ensure goal-concordant care of patients with CSLs who undergo a procedure. Efforts should be made to incorporate these guidelines into the training of all periprocedural providers.
BACKGROUND: Palliative care consultation during serious life-limiting illness can reduce symptom burden and improve quality of care. However, quantifying the impact of palliative care is hindered by the limitations of manual chart review and administrative coding.
OBJECTIVES: Using novel natural language process (NLP) techniques, we examined associations between palliative care consultations and performance on nationally endorsed metrics for high-quality end-of-life (EOL) care in patients with leptomeningeal disease (LMD) secondary to metastatic breast cancer.
METHODS: Patients with breast cancer with LMD were identified using administrative billing codes and NLP review of magnetic resonance imaging reports at 2 tertiary care centers between 2010 and 2016. Next, NLP was used to review clinical notes to (1) determine the presence of palliative care consultations and (2) determine the performance of process measures associated with high-quality EOL care, including discussions of goals of care, code status limitations, and hospice. Associations between palliative care consultation and documentation of EOL process measures were assessed using logistic regression.
RESULTS: We identified 183 cases of LMD. Median age was 56 (interquartile range [IQR]: 46-64) years and median survival was 150 days (IQR: 67-350). Within 6 months of diagnosis, 88.5% of patients had documentation of =1 process measure, including discussions of goals of care (63.4%), code status limitations (62.8%), or hospice (72.1%). Palliative care consultation was a predictor of subsequent documentation of goals of care (odds ratio [OR], 3.15; 95% confidence interval [CI], 1.58-6.27) and hospice discussions (OR, 4.61; 95% CI, 2.12-10.03).
CONCLUSION: Palliative care involvement is associated with increased performance of EOL process measures in patients with breast cancer with LMD.
Background: Natural language processing (NLP), a form of computer-assisted data abstraction, rapidly identifies serious illness communication domains such as code-status confirmation and goals of care (GOC) discussions within free-text notes, using a codebook of phrases. Differences in the phrases associated with palliative care for patients with different types of illness are unknown.
Objective: To compare communication of code-status clarification and GOC discussions between patients with advanced pancreatic cancer undergoing palliative procedures and patients admitted with life-threatening trauma.
Design: Retrospective cohort study.
Setting/Subjects: Patients with in-hospital admissions within two academic medical centers.
Measurements: Sensitivity and specificity of NLP-identified communication domains compared with manual review.
Results: Among patients with advanced pancreatic cancer (n = 523), NLP identified code-status clarification in 54% of admissions and GOC discussions in 49% of admissions. The sensitivity and specificity for code-status clarification were 94% and 99% respectively, while the sensitivity and specificity for a GOC discussion were 93% and 100%, respectively. Using the same codebook in patients with life-threatening trauma (n = 2093), NLP identified code-status clarification in 25.9% of admissions and GOC discussions in 6.3% of admissions. While NLP identification had 100% specificity, the sensitivity for code-status clarification and GOC discussion was reduced to 86% and 50%, respectively. Adding dynamic phrases such as “ongoing discussions” and phrases related to “family meetings” increased the sensitivity of the NLP codebook for code status to 98% and for GOC discussions to 100%.
Conclusions: Communication of code status and GOC differ between patients with advanced cancer and those with life-threatening trauma. Recognition of these differences can aid in identification in patterns of palliative care delivery.
BACKGROUND: Given survival measured in months, metrics, such as 30-day mortality, are poorly suited to measure the quality of palliative procedures for patients with advanced cancer. Nationally endorsed process measures associated with high-quality PC include code-status clarification, goals-of-care discussions, palliative-care referral, and hospice assessment. The impact of the performance of these process measures on subsequent healthcare utilization is unknown.
METHODS: Administrative data and manual review were used to identify hospital admissions with performance of palliative procedures for advanced pancreatic cancer at two tertiary care hospitals from 2011 to 2016. Natural language processing, a form of computer-assisted abstraction, identified process measures in associated free-text notes. Healthcare utilization was compared using a Cox proportional hazard model.
RESULTS: We identified 823 hospital admissions with performance of a palliative procedure. PC process measures were identified in 68% of admissions. Patients with documented process measures were older (66 vs. 63; p = 0.04) and had a longer length of stay (9 vs. 6 days; p < 0.001). In multivariate analysis, patients treated by surgeons were less likely to have PC process measures performed (odds ratio 0.19; 95% confidence interval 0.10–0.37). Performance of PC process measures was associated with decreased healthcare utilization in a Cox proportional hazard model.
CONCLUSIONS: PC process measures were not performed in almost one-third of hospital admissions for palliative procedures in patients with advanced pancreatic cancer. Performance of established high-quality process measures for seriously ill patients undergoing palliative procedures may help patients to avoid burdensome, high-intensity care at the end-of-life.
Few studies have investigated palliative and end-of-life care processes among young adults (YAs), aged 18-34 years, who died of cancer. This retrospective study used a natural language processing algorithm to identify documentation and timing of four process measures in YA cancer decedents' medical records: palliative care involvement, discussions of goals of care, code status, and hospice. Among 2878 YAs, 138 had a recorded date of death. In this group, 54.3% had at least one process measure documented early (31-180 days before death), 18.0% had only late documentation of process measures (0-30 days), and 27.5% had none documented.
BACKGROUND: Timely documentation of care preferences is an endorsed quality indicator for seriously ill patients admitted to intensive care units. Clinicians document their conversations about these preferences as unstructured free text in clinical notes from electronic health records.
AIM: To apply deep learning algorithms for automated identification of serious illness conversations documented in physician notes during intensive care unit admissions.
DESIGN: Using a retrospective dataset of physician notes, clinicians annotated all text documenting patient care preferences (goals of care or code status limitations), communication with family, and full code status. Clinician-coded text was used to train algorithms to identify documentation and to validate algorithms. The validated algorithms were deployed to assess the percentage of intensive care unit admissions of patients aged >=75 that had care preferences documented within the first 48 h.
SETTING/PARTICIPANTS: Patients admitted to one of five intensive care units.
RESULTS: Algorithm performance was calculated by comparing machine-identified documentation to clinician-coded documentation. For detecting care preference documentation at the note level, the algorithm had F1-score of 0.92 (95% confidence interval, 0.89 to 0.95), sensitivity of 93.5% (95% confidence interval, 90.0% to 98.0%), and specificity of 91.0% (95% confidence interval, 86.4% to 95.3%). Applied to 1350 admissions of patients aged >=75, we found that 64.7% of patient intensive care unit admissions had care preferences documented within the first 48 h.
CONCLUSION: Deep learning algorithms identified patient care preference documentation with sensitivity and specificity approaching that of clinicians and computed in a tiny fraction of time. Future research should determine the generalizability of these methods in multiple healthcare systems.
BACKGROUND: Palliative surgical procedures are frequently performed to reduce symptoms in patients with advanced cancer, but quality is difficult to measure.
OBJECTIVE: To determine whether natural language processing (NLP) of the electronic health record (EHR) can be used to (1) identify a population of cancer patients receiving palliative gastrostomy and (2) assess documentation of end-of-life process measures in the EHR.
DESIGN/SETTING: Retrospective cohort study of 302 adult cancer patients who received a gastrostomy tube at a single tertiary medical center.
MEASUREMENTS: Sensitivity and specificity of NLP compared to gold standard of manual chart abstraction in identifying a palliative indication for gastrostomy tube placement and documentation of goals of care discussions, code status determination, palliative care referral, and hospice assessment.
RESULTS: Among 302 cancer patients who underwent gastrostomy, 68 (22.5%) were classified by NLP as having a palliative indication for the procedure compared to 71 patients (23.5%) classified by human coders. Human chart abstraction took >2600 times longer than NLP (28 hours vs. 38 seconds). NLP identified the correct patients with 95.8% sensitivity and 97.4% specificity. NLP also identified end-of-life process measures with high sensitivity (85.7%-92.9%,) and specificity (96.7%-98.9%). In the two months leading up to palliative gastrostomy placement, 20.5% of patients had goals of care discussions documented. During the index hospitalization, 67.7% had goals of care discussions documented.
CONCLUSIONS: NLP offers opportunities to identify patients receiving palliative surgical procedures and can rapidly assess established end-of-life process measures with an accuracy approaching that of human coders.
BACKGROUND: Alone, administrative data poorly identifies patients with palliative care needs.
OBJECTIVE: To identify patients with uncommon, yet devastating, illnesses using a combination of administrative data and natural language processing (NLP).
DESIGN/SETTING: Retrospective cohort study using the electronic medical records of a healthcare network totaling over 2500 hospital beds. We sought to identify patient populations with two unique disease processes associated with a poor prognosis: pneumoperitoneum and leptomeningeal metastases from breast cancer.
MEASUREMENTS: Patients with pneumoperitoneum or leptomeningeal metastasis from breast cancer were identified through administrative codes and NLP.
RESULTS: Administrative codes alone resulted in identification of 6438 patients with possible pneumoperitoneum and 557 patients with possible leptomeningeal metastasis. Adding NLP to this analysis reduced the number of patients to 869 with pneumoperitoneum and 187 with leptomeningeal metastasis secondary to breast cancer. Administrative codes alone yielded a 13% positive predictive value (PPV) for pneumoperitoneum and 25% PPV for leptomeningeal metastasis. The combination of administrative codes and NLP achieved a PPV of 100%. The entire process was completed within hours.
CONCLUSIONS: Adding NLP to the use of administrative codes allows for rapid identification of seriously ill patients with otherwise difficult to detect disease processes and eliminates costly, tedious, and time-intensive manual chart review. This method enables studies to evaluate the effectiveness of treatment, including palliative interventions, for unique populations of seriously ill patients who cannot be identified by administrative codes alone.
STUDY OBJECTIVE: Do-not-resuscitate (DNR) status has been shown to be an independent risk factor for mortality in the post-operative period. Patients with DNR orders often undergo elective surgeries to alleviate symptoms and improve quality of life, but there are limited data on outcomes for informed decision making.
DESIGN: Retrospective cohort study.
SETTING: A multi-institutional setting including operating room, postoperative recovery area, inpatient wards, and the intensive care unit.
PATIENTS: A total of 566 patients with a DNR status and 316,431 patients without a DNR status undergoing elective procedures using the American College of Surgeons National Surgical Quality Improvement Program (ACS NSQIP) from 2012.
INTERVENTIONS: Patients undergoing elective surgical procedures.
MEASUREMENTS: We analyzed the risk-adjusted 30-day morbidity and mortality outcomes for the matched DNR and non-DNR cohorts undergoing elective surgeries.
MAIN RESULTS: DNR patients had significantly increased odds of 30-day mortality (OR 2.51 [1.55–4.05], p < 0.001) compared with non-DNR patients. In the DNR versus non-DNR cohort there was no significant difference in the occurrence of a number of 30-day complications, the rate of resuscitative measures undertaken, including cardiac arrest requiring CPR, reintubation, or return to the OR. The most common complications in both DNR and non-DNR patients undergoing elective procedures were transfusion, urinary tract infection, reoperation, and sepsis. Finally, the DNR patients had a significantly increased total length of hospital stay (7.65 ± 9.55 vs. 6.87 ± 9.21 days, p = 0.002).
CONCLUSIONS: DNR patients, as compared with non-DNR patients, have increased post-operative mortality but not morbidity, which may arise from unmeasured severity of illness or transition to comfort care in accordance with a patient's wishes. The informed consent process for elective surgeries in this patient population should include a discussion of acceptable operative risk.
Palliative surgical procedures are often performed for patients with limited survival. Quality measures for processes of care at the end of life are appropriate in palliative surgery, but have not been applied in this patient population. In this paper, the authors propose 4 quality measures for end-of-life care in a palliative surgery, and then demonstrate the utility of natural language processing for implementing these measures.
BACKGROUND: As our population ages and the burden of chronic illness rises, there is increasing need to implement quality metrics that measure and benchmark care of the seriously ill, including the delivery of both primary care and specialty palliative care. Such metrics can be used to drive quality improvement, value-based payment, and accountability for population-based outcomes.
METHODS: In this article, we examine use of the electronic health record (EHR) as a tool to assess quality of serious illness care through narrative review and description of a palliative care quality metrics program in a large healthcare system.
RESULTS: In the search for feasible, reliable, and valid palliative care quality metrics, the EHR is an attractive option for collecting quality data on large numbers of seriously ill patients. However, important challenges to using EHR data for quality improvement and accountability exist, including understanding the validity, reliability, and completeness of the data, as well as acknowledging the difference between care documented and care delivered. Challenges also include developing achievable metrics that are clearly linked to patient and family outcomes and addressing data interoperability across sites as well as EHR platforms and vendors. This article summarizes the strengths and weakness of the EHR as a data source for accountability of community- and population-based programs for serious illness, describes the implementation of EHR data in the palliative care quality metrics program at the University of Washington, and, based on that experience, discusses opportunities and challenges. Our palliative care metrics program was designed to serve as a resource for other healthcare systems.
DISCUSSION: Although the EHR offers great promise for enhancing quality of care provided for the seriously ill, significant challenges remain to operationalizing this promise on a national scale and using EHR data for population-based quality and accountability.