[ad_1]
Discussion
In this data-driven quantitative analysis, quality measures varied in their existence, value and distribution, even after normalisation, and provided no consistent insights across different integration analyses. In other words, we could not integrate quality measures to describe how the underlying quality constructs of interest—ICU visits, ED visits, palliative care use, lack of hospice, recent hospice, use of life sustaining therapy, chemotherapy and advance care planning—are used relative to each other across patients. Our findings suggest that quality measure calculations lack any interconnection information that can be potentially used to construct a story to provide insights about where, when or what care is provided to patients that can be used to support quality-improvement decision making. This is possibly for two reasons, (1) quality measure calculations mathematically abstract out siloed aspects of care across a cohort of patients, and by doing so, (2) disregard patient-level healthcare utilisation trajectories that contain the longitudinal interconnection between different services delivered. And yet, we posit that the administrative claims data—used to generate these quality measures—actually do contain such interconnection information because claims data include patient-level healthcare trajectories that include sequence order and timing for the delivery of different services at different places. Viewing healthcare systems from a lens of ‘general systems theory’ helps to explain why examining the parts of the system independently does not provide a systemic view of the behaviour of care delivery since emergent behaviours arise from the system that do not appear in any individual component.
Interconnection information can be elucidated by understanding the sequential utilisation of care by patients. For example, the interconnection between ICU visits and Life Sustaining Therapy can be examined by exploring their sequential utilisation across patients. Does Life Sustaining Therapy only occur following an ICU visit, or does it occur following an ED or inpatient (hospital) visit? Does Life Sustaining Therapy occur after all, most, some or no ICU visits? Similarly, the interconnection between Life Sustaining Therapy and Palliative care can be elucidated by examining their sequential utilisation. Interconnection information from sequential utilisation across multiple sets of quality constructs can then be used to develop insights about the behaviour of care delivered within a healthcare organisation that can be used to provide more actionable information about potentially how, where, when and for which patients to intervene to improve quality for multiple quality measures at once. Quality improvement decision making that takes into account multiple aspects of care quality, may also help alleviate unintended consequences that may arise from focusing on a single measure for improvement without reflecting on its impact on other measures, that is, other aspects of care. For example, Sedhom et al describe that in an effort to provide patient-concordant wishes of dying at home and not in a hospital, actions taken resulted in more patients dying not only at home but also in nursing facilities. For older patients, this action to avert a hospital death has led to the unintended consequences of only a 25% chance of dying at home and a much higher chance of dying in a nursing facility, where less attention is provided to symptoms, existential distress and grief compared with patients that are able to remain in their home.28
In quality measurement calculations, patients’ real-life sequential utilisation of care is ignored and only the relevant quality construct is extracted for each quality measure calculation independently. For each quality measure, the relevant quality construct (ICU use, palliative care use and so on) is aggregated across patients, essentially providing a siloed organisation perspective of quality that is disconnected from the patient’s perspective of quality. This calculation has two critical limitations. First, it does not distinguish, how many and which patients contributed to each quality measure—information which is insightful for quality improvement decision making. For example, if quality measures for ICU visits, ED visits and palliative care were ‘poor’, it is unclear if that is a result of different patients contributing to each of these quality measures independently, or if it is the same patients contributing to all these quality measures. In the former, quality improvement efforts may target different patients with different interventions, whereas, in the latter, a single intervention for all patients may ameliorate poor quality care for all measures. Second, quality measure calculations fundamentally ignore the temporal contiguity of care that can provide insights as to how patients are trajecting through the healthcare system both in terms of where they receive care (ICU, ED and so on) and what care they receive (palliative, life-sustaining therapy and so on). Instead, by aggregating patient-level trajectory information, underlying care patterns can emerge if patterns exist across patients.29 Hofstede et al have shown the importance of patient-level data, and how ecological fallacy potentially influences the interpretation of hospital performance when patient-level associations are not taken into account.30 Indeed, temporal patterns of care across patients can provide contextual insights about how, when, where and what care is delivered and, consequently, provide potential information as to how, when or where to intervene. Khayal et al have developed new systemic mathematical constructs (images) designed to convey interconnection information from administrative claims data to provide quality information in a new mathematical construction that goes beyond instances of many single quality measures.29
General systems theory describes that for systems, such as healthcare systems, with interconnected components, emergent behaviours arise from the system that does not appear in any individual components.31 For cancer care delivery, and in the typical case of a patient receiving care from multiple specialties across several settings, a change in a healthcare provider’s decisions and actions can change the context for other healthcare providers. This phenomenon of complex systems explains why an examination of the parts, such as through quality measures, gives us no information about the coordination of parts and processes.30 In other words, it is in the coordination and interconnection of information between the parts or processes that higher-level behaviours emerge. It is those behaviours that need to be understood to develop quality improvement interventions that can target them. While very few examples exist of using system-based methodologies such as discrete-event simulation, agent-based modelling and others to develop an understanding of the behaviour of a system from administrative claims data,29 it is fundamentally in the aggregation of the sequence of delivered care (patient healthcare trajectory) where patterns of systemic interconnection can emerge. Consequently, a systems approach to quality measurement—with the creation of systemic mathematical constructs—is very likely required to make system-level improvement decisions that take into consideration upstream and downstream effects to minimise unintended consequences.
[ad_2]
Source link
