The Algorithms of Death: RESPECT: An Algorithm for Predicting Time to Death for 'Frail Older Adults'

 

Pix Credit HERE

 

Predictive analytics has become an indispensable tool of markets, and of the state. As the role of both shifts from structural to managerial responsibilities--from creating and maintaining frameworks within which individuals in collectives can order their lives and make choices to  transforming frameworks into  systems of nudging to induce individuals to make the "right" choices--the need for data based analytics becomes more acute.  These analytics firsts help those whose business it is to ensure appropriate conduct to understand the collective behaviors of the target population.  Second, on the basis of that understanding, it is then possible to predict behavior or outcomes given  functional relationships among the factors that are critical to the model.  Eventually, a sophisticated enough analytics (what some might also call algorithms, though the algorithm is better understood as the calculation of judgment derived from the application of an analytics either to describe or predict) can be used to control behavior by changing the intensity or direction of factors (coefficients of values) that produce behavior movement in particular directions. 

However, one cannot get to the third stage (the ultimate objective of planners and those seeking to mold society toward ideal types) without a sound basis in predictive analytics.  It is to that end that modelers are now focusing with greater intensity on predicting all sorts of behaviors, conditions, events, and effects that touch on regulatory, policy, or market roles for societal or governmental collectives or institutions. Not that there is some sort of sinister object at work here.  The move from qualitative to quantitative and comprehensive management of populations has been a long time coming and has emerged in strong alignment with cultural, political, and societal changes that have appeared to increase (as technology makes possible) an appetite  for the scientific management of life in organized society--even (or especially in live within the liberal democratic  collectives of states).  At some point, there is no doubt, our sociologists (if permitted--for by then it may be a taboo to study these things) will be able to explain this draft and then rush toward  administrative managerialism within both public and private sector components of liberal democratic societies. Foe the moment it may be enough to knwo that it appears now to be an irresistible force--one augmented by the promise of technology and the reassurance of certainty in collectives making decisions that affect the lives of individuals in ways that are said to align with collective societal (or political) desire.

 It is in that context that one might usefully encounter an important manifestation of that trend in the so-called RESPECT project that has recently garnered the voyeuristic attention of the press (though hardly its understanding of other it or its context). .

The algorithm used in this calculator is called RESPECT. RESPECT is short for Risk Evaluation for Support: Predictions for Elder life in the Community Tool.

RESPECT was developed on data from 491,277 older people in Ontario, Canada, who used home care between 2007 and 2013. The dataset contains detailed health information from the standardized Resident Assessment Instrument for Home Care (RAI-HC), which case managers and nurses use to assess the needs of home care users. The calculator includes a wide range of questions that reflect different trajectories in physical health and cognition, including age, sex, cognitive impairment (memory decline), diseases (e.g. diabetes, cardiovascular disease, dementia, cancer), sociodemographic factors (marital status, level of education), health status (e.g. difficulties with activities of daily living), symptoms of reduced physiologic reserve (e.g. weight loss), use life-sustaining therapies (e.g., dialysis), and health care use (e.g., number of hospitalizations and emergency department visits).

The algorithm is calibrated to 1.3 million assessments and 112,823 deaths. RESPECT accurately predicts a wide range in life expectancy; we ranked and classified our cohort based on their six-month risk of death, which spanned from 1.5% to 98%. This translates to an average survival of 4 weeks (interquartile range [IQR] of 11 to 84 days) for the frailest people in our cohort to 8.1 years among people with the few physical limitations and chronic health conditions (IQR of 5.9 years to 9.4 years).

RESPECT is incorporated into the Project Big Life scoring engine. The scoring engine generates calculations on any combination of responses to the algorithm questions. The main output of the engine is the actual health experience of Ontarians who completed the 1.3 million assessments. These health experiences can be translated into a wide range of patient-oriented measures. For more information on how we developed the RESPECT–End of Life algorithm, click here. (RESPECT website HERE))

RESPECT was developed by a team of researchers Amy T. Hsu, Douglas G. Manuel, Sarah Spruin, Carol Bennett, Monica Taljaard, Sarah Beach, Yulric Sequeira, Robert Talarico, Mathieu Chalifoux, Daniel Kobewka, Andrew P. Costa, Susan E. Bronskill and Peter Tanuseputro, and recently published CMAJ 193(26):E997-E1005 (5 July 2021) under the title: Predicting death in home care users: derivation and validation of the Risk Evaluation for Support: Predictions for Elder-Life in the Community Tool (RESPECT) (ARTICLE PDF HERE)

The article is important reading for anyone interested in the forward movement of modeling, and of its potential as a tool in a variety of contexts and with a variety of consequences.  My very brief observations follow along with the "Interpretation" and "Methods" Sections of that article, and the FAQs from its Website. the CALCULATOR may be accessed HERE.

 

 

1. The work suggests the increasing tendency to build models on models (I have spoken to that phenomenon HERE).  Thus for example, the authors noted that "selection of predictors included in our model was informed by clinical experience and our review of existing mortality prediction models for a population of older, community-dwelling adults." The need to model from models is both necessary but also increases those points of risk for the robustness of the predictive model.  Predictive analytics are increasingly based on ecologies of sub modelling. Models, the,selves assume dynamic and changeable character--they are outcomes ( the analytic expression of  a rationalized relationship among factors duly weighted in accordance with ideology and objectives) and they are consumables (the object-components of larger models). 

2.  The product of these models of models--like this death predictor--then itself becomes both a judgment and an object that can be used for decisionmaking in both pubic and private institutions.  Of course, the objective that drove its creation was no doubt for the best of reasons--in this case to treat end of life more humanely. But that is hardly the end of it.  The data of course acquires a life of its own.  And there is always the temptation of tracking subjects.  More important, of course, would be the drive to aligning simulation (a portion fo the population to sand in for the whole) to the whole of the population istelf; where the model merges with is subject.  The value of the algorithm may substantially impact the impact o such a move. 

3. Moreover, the data generated lends itself to multiple uses.  It would be quite tempting (and we speak here of temptations in multiple sectors and among multiple parties--so great is the potential of algorithms like this one) for example for the financial industry to insist (or to find ways of developing their own mirroring metrics) on the use of this predictive analytics as an aid for everything for pricing coverage (insurance) to rating creditworthiness. Here law will be slow to follow the likely transformation of private practices where such actors may get access to the analytics (or more richly) the data.  

Pix Credit HERE

4. Closer to home the medical establishment might make uses of this beyond the  imagination of its creators.  For example, decisions about and the construction of medical protocols may well become a function of the results of the application of these metrics. One moves here from the soft intuitions of actuaries, to the more solid predictors of the algorithm.  Where in crisis resources are limited, one might be able to combine choices among appropriate medical interventions with value choices about hierarchies of treatment. The medical establishment puts the choices nicely though what they are suggesting is the way in which this algorithm can substitute their judgment, or societal values of the extent of aggressive treatment: "'Knowing how long a person has to live is essential in making informed decisions about what treatments they should get and where they should get them,' says Dr. Peter Tanuseputro, physician-scientist at The Ottawa Hospital and ICES, and investigator at the Bruyère Research Institute. "As a person gets closer to death, the balance shifts from having curative care as the primary goal, to care that maximizes a person's quality of remaining life.'" (New online calculator can help predict death and end-of-life care needs for older adults).

5. The ultimate object is efficiency--efficient allocation of resources, efficient diagnosis, efficient evaluation of risk and the like. The efficiency objective, of course, avoids the policy question that the analytics cannot answer--but to which the algorithmic judgments must be held accountability, and that is the values embedded in notions of efficiency in the variety of contexts in which it applies. Efficiency is not merely a quantifiable metaphor; it is the end product of a complex application of valuation that is itself a translation of normative judgments--normative judgments that are themselves complex comparative pricing systems.

6. The model was not unconsciously projected onto contemporary normative debates--but by offering a quantitative overlay on complex policy and normative issues that ought to determine its construction and use.  It is not for nothing that its authors appeared to have worked hard on its acronym: RESPECT. But their is irony here as well because the RESPECT implies a direction that may be belied by its use: respect for resources, respect for the aged, respect for the integrity of medical establishments, respect for the state as the arbiter of  support for life at its end.

7. Ultimately, though, what one sees possibly emerging here is the "scientific" (or rationalized) allocation of risk (or the reduction of risk) by those for whom mortality matters. The algorithm, like much of analytics, is alluring precisely because it shifts the burden of risk for decisionmaking from individuals to the analytics itself. And because in our society people are taught that there is no arguing about "numbers" not only is risk reduced among individuals with decision making authority (by shifting 'blame' to a 'formula' but then effectively creates an immunity in the formula itself (other than for negligence in its application).  In a society increasingly embracing the fundamental ordering principle of risk avoidance--the algorithm and its analytics provides the perfect vehicle for embracing risk for those who must bear it while reducing or shifting risk from those who now control it (and the algorithm that serves as the risk allocation platform). But because the 'politics' of the algorithm are veiled in the language of logic, mathematics and modeling, it becomes much harder for the population which will feel its effects to effectively engage with its premises, applications, or uses.

 

__________

Amy Hsu et al  Predicting death in home care users: derivation and validation of the Risk Evaluation for Support: Predictions for Elder-Life in the Community Tool (RESPECT) CMAJ 193(26):E997-E1005 (5 July 2021)

* * *  

INTERPRETATION: The RESPECT mortality risk prediction tool that makes use of readily available information can improve the identification of palliative and end-of-life care needs in a diverse older adult population receiving home care.

Most people in high-income countries die of causes with progressive, predictable trajectories of decline.1^–4 Since 2000, the 3 leading causes of death in Canada — accounting for 55% of all deaths — have been cancer, heart disease and stroke.1 Other leading causes of death, such as dementia and chronic lower respiratory diseases, also share signs and symptoms of senescence that are common across chronic diseases, including deterioration of physical and cognitive function, as well as an increased need for assistance.

Despite the predictable nature of most deaths, many Canadian residents who are at the end of life do not receive adequate home-based supports.5 In Ontario — the largest province in Canada with more than 14 million residents and the setting of this study — only 40% of decedents receive formal home care, and less than 20% receive a physician home visit in their last year of life.6^,7 Even among those who had received palliative and end-of-life care, the start of service was often too close to death and failed to have a positive impact on the quality of life in those last months.8 The lack of available and accurate prognostic information is a key challenge. There are few existing tools that can be used to inform palliative care planning for the general population of older adults who live in the community and in people without cancer.9 Other barriers to accurate prognostic estimates include clinicians’ reluctance or lack of time and existing prognostication tools’ reliance on complex or specialized inputs, such as laboratory data and previous health care use. As a result, many older and frail adults do not receive timely palliative care and do not have an advance care plan.6^,10^–13

Our primary objective was to develop and validate a model for predicting mortality risk among the general population of community-dwelling adults with and without cancer that spans an actionable period for end-of-life planning (5 yr to imminent death). The variables included in our prognostication model — the Risk Evaluation for Support: Predictions for Elder-life in the Community Tool (RESPECT) — were prespecified to include exposures that could be self-reported by patients and their caregivers, including family members.

Methods

We derived and validated RESPECT using population-based home care data housed at ICES in Ontario. The reporting of our approach and findings adhere to the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) statement.14 A detailed protocol prespecifying our approach was published elsewhere.15

Study population

The study population included people who were 50 years of age or older, eligible for government-funded long-term home care in the community and had received at least 1 assessment using the Resident Assessment Instrument for Home Care (RAI-HC) between Jan. 1, 2007, and Dec. 31, 2013. The derivation cohort included those who used home care between Jan. 1, 2007, and Dec. 31, 2012, and the temporal validation cohort received care and were assessed between Jan. 1 and Dec. 31, 2013. The RAI-HC is a comprehensive, multidimensional instrument for clinical assessment that contains nearly 400 data elements that capture the home care client’s sociodemographic profile, cognitive and functional capacities, chronic diseases and comorbidities, and signs of health instability, as well as recent use of health care. It is primarily used for care planning within the home care setting. However, it also contains quality indicators and outcome measures that could be used to evaluate the impact of services provided. In Ontario, home care services, such as nursing, personal support and rehabilitative or restorative therapy, are publicly funded and provided to people to improve or manage their health based on assessed need. The RAI-HC is used by home care coordinators to evaluate the level of care need in those who are expected to require at least 60 days of uninterrupted service (also known as “long-stay clients”). Evaluations using the RAI-HC are performed at initial consideration for home care and completed at least once every 6 months for those receiving home care over an extended period, or when substantial changes in the client’s situation have been observed.

The primary outcome of our prediction model was death within 6 months of a RAI-HC assessment. Death was ascertained from the Registered Persons Database, a registry of health card numbers that have been issued under the Ontario Health Insurance Plan to all eligible residents of Ontario.

The selection of predictors included in our model was informed by clinical experience and our review of existing mortality prediction models for a population of older, community-dwelling adults. We also considered variables included in existing frailty indices.9 We considered risk factors related to physical functioning (e.g., difficulties with activities of daily living [ADL], inability to independently carry out instrumental activities of daily living [IADL] and reduced mobility), cognitive impairment (e.g., memory decline and psychosis), sociodemographic factors (e.g., level of education) and biological diseases (e.g., diabetes, heart disease and cancer). We also included self-reported measures of recent use of health care (i.e., number of hospital admissions or visits to the emergency department in the last 90 d), the prescription and receipt of life-sustaining therapies (e.g., dialysis and ventilation), and symptoms of reduced health and physiologic reserve (e.g., weight loss, edema and vomiting). The latter variables were selected to capture the acute symptomatology of people in the terminal period of life. We also incorporated cohort characteristics (e.g., year of the RAI-HC assessment, and the type of and reason for assessment) that may account for remaining heterogeneity in the estimated risks. No stepwise variable selection procedure was used.

 

__________

 

Frequently Asked Questions

What is RESPECT?

RESPECT is short for Risk Evaluation for Support: Predictions for Elder-life in the Community Tool. It was developed for frail older people who might need supports and care in their homes. Technically, it is an algorithm that calculates a frail person’s survival—that is, how long they will live. Using the responses to 17 questions about their health and ability to care for themselves, the tool provides an estimate of a person’s survival based on information gathered on people who have similar characteristics.

Who will use RESPECT?

RESPECT was designed for frail older adults living in the community who are uncertain about their survival and the family members and caregivers who support them. RESPECT can also be used by formal care providers—such as physicians, home care staff or palliative care teams—to understand their patient’s decline.

How will RESPECT be used?

As a person’s health declines, they may need more supports and care in their home. RESPECT calculates a person’s survival and provides information that can help them understand what type of care and services they may need. A patient can use this information to discuss their care needs with their caregivers and healthcare providers. Similarly, formal care providers can use this tool to discuss, with their patient, what can be expected as the patient approaches the end of life and plan for the supports that their patient may need.

How was RESPECT created?

RESPECT was developed and validated using home care data from Ontario. We used data from 1.3 million Ontario home care assessments that were followed to 80,000 deaths between 2007 and 2014. The dataset contains detailed health information from the standardized Resident Assessment Instrument for Home Care, which case managers or care coordinators commonly use to assess the needs of frail individuals who might need home care.

Can RESPECT be used by those not receiving home care?

RESPECT was designed for frail older adults living in the community, which includes those living in assisted living and retirement home communities, who might need support at home. People receiving care in other settings—such as those in long-term care or nursing homes—may have different underlying risks that contribute to their survival and RESPECT will not be able to accurately estimate the mortality risk for these populations.

Who developed RESPECT?

RESPECT was created by the Project Big Life Team, which includes patients and caregivers to older adults who need care. The Project Big Life Team is led by researchers at the Ottawa Hospital, the Bruyère Research Institute, the University of Ottawa and ICES. ICES, formerly known as the Institute for Clinical Evaluative Sciences, is a not-for-profit research institute encompassing a community of research, data and clinical experts, and a secure and accessible array of Ontario's health-related data.

Can RESPECT be integrated with electronic medical records?

Through funding from the Canadian Institutes of Health Research, the Project Big Life Team will be collaborating with various healthcare providers and organizations to improve the integration of this information into existing electronic medical records. At this time, results generated from the online calculator can be printed and shared with a patient’s formal care provider and those in their circle of care.

Who funded this research?

This work was funded by the Canadian Institutes of Health Research and supported by ICES. ICES receives core funding from the Ontario Ministry of Health and Long-Term Care.

.