Quality of Palliative Care for Patients With Advanced Cancer in a Community Consortium

Abstract

Background

Measuring quality of care delivery is essential to palliative care program growth and sustainability. We formed the Carolinas Consortium for Palliative Care and collected a quality data registry to monitor our practice and inform quality improvement efforts.

Measures

We analyzed all palliative care consultations in patients with cancer in our quality registry from March 2008 through October 2011 using 18 palliative care quality measures. Descriptive metric adherence was calculated after analyzing the relevant population for measurement.

Intervention

We used a paper-based, prospective method to monitor adherence for quality measures in a community-based palliative care consortium.

Outcomes

We demonstrate that measures evaluating process assessment (range 63-100%), as opposed to interventions (range 3-17%), are better documented.

Conclusions/Lessons Learned

Analyzing data on quality is feasible and valuable in community-based palliative care. Overall, processes to collect data on quality using non-technology methods may underestimate true adherence to quality measures.

Background

Health care systems are currently evolving to meet the triple aims for highly effective health care proposed by the Institute for Healthcare Improvement (IHI), including delivering high-quality, low-cost care to large populations. To date, data collection efforts around quality have been limited by scope (e.g., analyses limited to institutional level), validity (e.g., retrospective, aggregated data parsed from administrative and billing databases), and clinical relevance (e.g., limited capability of data to impact ongoing patient care). As pay-for-performance models are phased in, quality measurement will be needed for regular demonstration of quality of care and improving patient outcomes. Consequently, the portfolio of quality measures to evaluate structure, process, and outcomes are rapidly expanding in all fields of medicine, including consultative palliative care.

With the dramatic growth in clinical palliative care services over the last decade comes the unique opportunity to collaborate, compare, and learn from each other. Indeed, new collaborations are forming, such as: the Palliative Care Research Cooperative Group,¹ the first American national clinical research network for palliative care; the Coalition of Hospices Organized to Investigate Comparative Effectiveness, a distributed hospice network providing data for comparison of quality of care and benchmarking; and several regional palliative care quality monitoring networks such as those in California and the Carolinas.^3,4 Multisite efforts require a transition towards information exchange and standardization of data in order to contribute to aggregate understanding.

The Carolinas Palliative Care Consortium is a novel academic/community collaboration established in 2007. ⁴ The vision is to “improve the quality of care of patients with advanced illness through benchmarking and quality initiatives using a data-driven system that monitors outcomes.” The resultant ongoing, collaborative venture has demonstrated capabilities for data collection and performance improvement.⁵ A novel technology environment supports efficient data collection at point of care. Herein, we present the first analysis of data collected regarding rates of adherence to quality measures, explore possible rationales for shortcomings, and signal opportunities for quality improvement projects.

Measures/Intervention

We included all patients with a cancer diagnosis (Table 1) between March 1, 2008 and October 1, 2011, in the Carolina Consortium Palliative Care Database,⁴ an Institutional Review Board-approved registry combining information from four member sites (Four Seasons, Flat Rock; Forsythe Medical Center, Winston-Salem; Hospice of Wake County and Horizons Palliative Care, Raleigh; and Charlotte Region Hospice and Palliative Care, Charlotte). Accompanying registry, data use, and business associates agreements govern the proper use and sharing of information.

Table 1.

Demographics of Patients With Cancer (N=459)

Variable	Categories	n	%
Age, yrs	<65	140	30.4
	≥65	303	65.9
	Missing	17	3.70
Gender	Male	204	44.4
	Female	239	52.0
	Missing	17	3.7
Race	African American	49	10.7
	Caucasian/White	385	83.7
	Other	3	0.65
	Missing	23	5.0
Cancer Type	Gastrointestinal	113	24.7
	Hematologic malignancies	24	5.5
	Lung	133	29.1
	Genitourinary	72	15.8
	Breast	39	8.5
	Other	76	16.6
	Missing	2	0.65
Life expectancy	“Hours to days” and “Days to weeks”	124	23.9
	“Weeks to months” and “4 to 6 months”	243	39.6
	Greater than 6 months	68	28.4
	Missing	25	5.43
PPS Level	10%-30%	182	40.0
	40%-60%	196	42.6
	70%-100%	36	8.0
	Missing	46	10.0

PPS=Palliative Performance Scale.

In order to generate the dataset, quality monitoring information was collected by palliative care clinicians at the point of care using the Quality Data Collection Tool – Palliative Care (QDACT-PC) v. 1.0, a palliative care-specific needs assessment tool developed by the Consortium. This tool incorporated provider-recorded and patient-reported data through multiple platforms of input, including paper and electronic pens. Demographic data, including patient age, sex, and race, were obtained from local administrative datasets and supplemented by clinician data entry. After patient identifiers and clinical data were matched locally, data were then transmitted electronically from local databases to the central registry and de-identified prior to analysis. Performance status (measured by the Palliative Performance Scale [PPS]⁶) and provider estimation of prognosis were available for analysis. To supply clinical data, clinicians entered the primary medical condition; diagnoses were confirmed by comparing with International Classification of Dieases-9 codes.

Adherence to 18 supportive or palliative care quality measures, derived from the American Society of Clinical Oncology Quality Oncology Practice Initiative, Hospice PEACE (Prepare, Embrace, Attend, Communicate, and Empower) Project, and Cancer-ASSIST quality metrics sets, was calculated^7-9 (Table 2). Measures were selected by members of the Consortium to reflect clinician interests and priorities. The methods outlined by the quality measure developers were used to calculate adherence, including designation of populations of interest in which the measure was intended to be assessed (denominators). A waiver of written consent was granted by the Duke University Institutional Review Board.

Table 2.

Adherence With Measures by Measure Type

Measure type/Source	Specific measure	Numerator/Denominator	Overall Adherence (%)
Assessment/PEACE	Percent of patients with chart documentation of their preference for life-sustaining treatments	459/459	100
Assessment/ Cancer-ASSIST	IF a cancer patient has a cancer-related outpatient visit THEN there should be screening for the presence or absence and intensity of pain using a numeric pain score.	459/459	100
Assessment/QOPI	Pain assessed by second office visit	454/459	99
Assessment/PEACE	Percent of patients screened for pain during the admission visit	453/459	99
Assessment/PEACE	Percent of patients who were screened for shortness of breath during the admission visit	453/459	99
Assessment/QOPI	Constipation assessed at time of narcotic prescription	390/459	85
Assessment/PEACE	Standard assessment for depression	376/459	82
Assessment/ Cancer-ASSIST	IF a cancer patient is seen for an initial visit or any visit while undergoing chemotherapy at a cancer-related outpatient site, THEN there should be an assessment of the presence or absence of fatigue.	368/459	78
Assessment/QOPI	Patient emotional well-being assessed by second office visit	289/459	63
Management/QOPI	Dyspnea addressed appropriately		17
Management/Cancer-ASSIST	IF an outpatient with primary lung cancer or advanced cancer reports new or worsening dyspnea, THEN s/he should be offered symptomatic management or treatment directed at an underlying cause within 1 month.	78/459	17
Management/QOPI	Plan of care for moderate/severe pain documented	78/459	16
Management/PEACE	For patients who screened positive for pain, the percent with any treatment within 1 day of screening	41/459	9
Management/PEACE	For patients who screened positive for dyspnea, the percent who receive treatment within 1 day of screening	28/459	6
Management/PEACE	For patients who screen positive for constipation, the percent who receive treatment within 1 day of screening	18/459	4
Management/PEACE	For patients who screen positive for depression, the percent who receive further assessment, counseling or medication treatment	14/459	3
Management/Cancer-ASSIST	IF depression is diagnosed in a cancer patient, THEN a treatment plan for depression should be documented.	13/459	3

Outcomes

From 2008 through 2011, 459 adult cancer patients were seen in the Carolinas Palliative Care Consortium. Patient age ranged from 18 to 106 years; 69% were aged 65 years or older. The majority of patients (85%) had a PPS less than 60% and 62% of patients had weeks to six months expected prognosis. Most patients had a diagnosis of lung cancer or a gastrointestinal malignancy (29% and 25%, respectively) (Table 1). Cancer stage and receipt of concurrent cancer-directed therapy were not recorded.

The distribution of adherence varied markedly (Table 2). Adherence was highest for those measures focused on assessment (63-100%). Only advance care planning, pain and dyspnea assessment were nearly uniformly conducted. Adherence dropped substantially for all measures focused on management. Measures of lowest adherence involved timely documentation of management for non-pain symptoms, especially depression.

Conclusions/Lessons Learned

Collaborative data sharing on quality is valuable in palliative care. We have demonstrated that collecting uninterrupted data, aided by paper and electronic methods in a community-based collaborative, can inform potential deficiencies and areas for performance improvement. Interestingly, our data demonstrate that measures aimed at distress assessment were more likely to be performed and documented than those addressing management. This finding is hypothesis-generating and has informed ongoing quality improvement efforts addressing proper and timely management of distress areas within our Consortium.

Noting the deficiencies in timely dyspnea and constipation management, we have recently developed and implemented a quick action response to foster performance improvement. We have based this approach on those of “rapid learning health care,”¹⁰ where data collected during usual care are cycled back into decision making. We have termed our efforts “rapid cycling quality improvement” where continuously collected data are fed back into a rapid quality improvement project. This approach aims to increase adherence to quality measures of interest in short time periods using directed feedback, comparisons, and a clear focus on a few measures at a time. Based on our results, the Consortium has focused on timely dyspnea and constipation management as the targets of ongoing performance improvement initiatives. We are aiming to improve our adherence to those quality measures to greater than 90%, while incorporating increasingly thorough audit and feedback reports to our providers.

It is difficult to explain why adherence is lower for some quality measures than others. We believe that low adherence, especially for measures that involve management, may be related to two documentation challenges posed by paper-based quality monitoring. First, management interventions for uncontrolled symptoms can often be difficult to choose simultaneous with performing the symptom assessment. As the QDACT tool was used during the provider’s interview with the patient, it may be that decisions on management were made later when the QDACT tool was already complete or submitted. For example, a provider may record a dyspnea score, based on the recorded score refer a patient for a spirometry test, and then because of demonstrated obstructive lung disease, start the patient on a long-acting bronchodilator. This decision may occur long after the QDACT record is transmitted to the registry. Second, the inherent limitations of paper-based data collection forms may limit the thoroughness of our registry. For example, queries that require a long, narrative answer may be incomplete because of time constraints faced by clinicians during patient encounters. Because of this, answers left blank may actually reflect the challenges of completing management questions at point-of-care, not the absence of an actual management decision.

Through our productive collaboration within the Carolinas Consortium, we have learned several key lessons. These include the value and importance of data sharing, the need to transition away from chart-based retrospective abstraction, and the potential to further define the logistics, processes, and methods for which quality should be collected, analyzed, and reported. This work is unique in that it analyzes patients from a community-based dataset, thus better characterizing the happenings outside of academic institutions. As palliative care grows, so too does the data within. This study of community-dwelling cancer patients broadens the current evidence that further improves the care for patients with advanced cancer. Ultimately, this will be necessary for the field of palliative care to stay in step with other medical disciplines that are also striving to keep up with the evolving landscape of quality, value, and reimbursement in the national discussion around health care reform.