Global Assessment of Palliative Care Need: Serious Health-Related Suffering Measurement Methodology

General considerations in the selection processes

The selection of conditions, development of multipliers, and calculation of the number of people and days of SHS was informed by a literature search, individual and group expert discussions, and Delphi processes with online surveys for SHS 1.0 as described in the Appendix to the Commission report. Expert panel(s) of palliative care clinicians with experience providing clinical care in different parts of the world, especially in LMICs were engaged in the process.

To estimate symptoms and symptom duration (days of SHS), as part of the work of the Commission and SHS1.0, experts were asked to consider a typical patient with each of the conditions and based on their daily experience, to generate an estimate of the prevalence and duration of each symptom. During the expert consultation stage, including focus group discussions and semi-structured interviews, results from the literature review were presented. Experts were asked to provide responses and feedback based on their work experiences even when those experiences were contrary to the evidence presented to them. Either individually or in groups, all data and estimates were vetted, considering assumptions and limitations or gaps to ensure that all relevant aspects or scenarios are reasonably accounted for when possible. It is expected that these data will serve to provide content validity for estimation of the global burden of remediable suffering.(14) See appendix table 1 for a full list of the experts’ consensus building practices undertaken by the LC.

Finally, the Delphi method for consensus-building also was used to determine the duration (average number of days requiring palliative care) for which palliative care was needed for each of the conditions included in the database.(15) Experts were purposively sampled and were considered to be ‘informed individuals’(16) and ‘specialists’(17) within the field of palliative care, in this case palliative care.(18) Both rounds of the Delphi requested 18 palliative care experts living in LMICs to estimate the number of days of palliative care that would be required for a patient with each of the given conditions. The responses from the first round were pooled to identify a group average range and standard deviation for each condition. The second round of the Delphi presented respondents with the average range of days of palliative care with confidence intervals for each parameter. Experts were asked to respond again to the same questions based on knowledge of the group’s prior responses. The response rate for round one was 83% and for round two was 27%. Results from each round are presented in Tables 2. See appendix table 2 for the results from the rounds 1 and 2 of the Delphi study.Due to limited resources, estimation for symptom-days is only available from the Lancet Commission (SHS1.0) and was not updated for SHS 2.0.

Selection of SHS-associated conditions

The first step in estimating the SHS burden was to identify the health conditions that most commonly cause SHS from the ICD-10 classification list that require palliative care at the end-of-life due to life-threatening conditions or living with a life-limiting condition (SHS 1.0). The global SHS 2.0 database includes 21 conditions, and these are presented in Table 1 with their corresponding ICD-10 codes and GBD codes. [Table 1: 21 conditions included in the global SHS database in ICD-10 codes and GBD disease codes] All 21 groups of conditions include decedent categories, considering that at least a proportion of people dying from those conditions suffer from serious health-related suffering. In addition, non-decedent categories of SHS are included for some of the 21 conditions that: may have been cured but from which SHS persists (drug-resistant tuberculosis, some hemorrhagic fevers such as Ebola, some malignancies, some inflammatory diseases of the central nervous system); from which patients recover but that caused SHS (serious injuries, renal failures, preterm birth complications and birth trauma); with survival with chronic severe disability and with SHS symptoms (cerebrovascular disease, leukemia, congenital malformations, injury, birth trauma, human immunodeficiency viruses / acquired immunodeficiency syndrome (HIV/AIDS), some musculoskeletal disorders, liver diseases); and, have a slowly progressive course (malignancies, dementia, Parkinson’s disease, multiple sclerosis, type-1 diabetes, thalassemia, and sickle cell disorders).

In the original Lancet Commission report, the non-decedents category for 11 conditions were considered. In SHS 2.0, non-decedents categories for four more conditions were added and differentiating factors were used that are important to estimating suffering patterns. Panel 1 provides a detailed description of how decedents and non-decedents in need of palliative care are estimated for each condition as well as key literature and extra databases used to calculate the decedents and non-decedents with SHS. Conditions are ranked using the alphabetical order of their ICD-10 codes.

As the result of the exercise to estimate palliative care needs for children, there was consensus that the following conditions be added due to their substantive contribution to SHS among children for both decedents and non-decedents: 1) diabetes mellitus, 2) sickle cell disorders, 3) thalassemia, and the following conditions for non-decedents categories of: 1) leukemia, 2) liver diseases, 3) chronic kidney diseases, 4) neonatal preterm birth and birth trauma. Hence, while SHS 1.0 included 20 conditions, SHS 2.0 includes 21 groups of conditions with the addition of endocrine, metabolic, blood, and immune disorders which include diabetes mellitus, sickle cell disorders, and thalassemia for decedents and non-decedents.

The review of the case of diabetes in children prompted an overall review of the included conditions. For diabetes mellitus in adults, deaths from sequelae are attributed to the proximal cause and hence considered captured in other conditions included in the SHS database and specifically, cerebrovascular disease, cardiomyopathy and/or heart failure, chronic ischemic heart disease, renal failure, and atherosclerosis. Because deaths from diabetic ketoacidosis or hyperglycemic hyperosmotic non-ketotic syndrome typically result in death so rapidly that there is no time to institute quality palliative care services, these conditions are not included. In the pediatric population, diabetes mellitus is added due to the concerns over pain and suffering caused by type-1 diabetes even in the absence of organ complications.

Efforts to alleviate SHS experienced by a newborn, the assurance of the newborn’s comfort and that of distraught parents should accompany aggressive life-sustaining treatments if they are to reasonably provide more benefit than burden. Palliative care must also be available as an alternative to potentially harmful life-sustaining interventions when a newborn is moribund. Hence, in both SHS 1.0 and SHS 2.0, extremely premature and very low birth weight newborns whose survival is unlikely, and babies born with severe hypoxic ischemic encephalopathy or congenital anomalies not compatible with life are included in the list of SHS conditions.

In both SHS 1.0 and SHS 2.0, leukemia is considered a separate condition than the rest of the malignancies due to its distinctive patients’ demographics and suffering patterns.

Selection of types or symptoms of suffering

Patients’ suffering varies by type, severity, and duration and a clinically, economically, and strategically useful measure of SHS requires estimation of not only the number of patients who suffer, but also the type of suffering and duration of suffering. Therefore, overarching categories of suffering were identified in SHS1.0 and then within those categories, the types or symptoms were associated with each condition.

Palliative care literature typically divides suffering into four categories – physical, psychological, social, and spiritual – to encompass the full spectrum of human suffering. While the Lancet Commission accepted and adopted all four categories as SHS, the focus was on estimating the prevalence and duration of only physical and psychological categories of suffering and corresponding symptoms. The empirical evidence from published literature or expertise to produce reasonable estimates of the prevalence and duration of each type of social and spiritual suffering were not sufficient.

To estimate SHS as precisely as possible, the Commissiońs expert group identified the most common symptoms of physical and psychological suffering, and then estimated the prevalence and duration of each type of suffering associated with each condition or its treatment. Through literature review and evidence-informed expert consensus building exercises, physical and psychological types of suffering (symptoms), their frequencies and durations for each condition were identified as part of Commission work. See [Figure 1] for details. Specifically, the types of physical suffering include: moderate or severe pain, mild pain, weakness, fatigue, shortness of breath, nausea and vomiting, constipation, diarrhea, dry mouth, itching, wounds and bleeding. The types of psychological suffering identified include: – anxiety and worry, depressed mood, delirium or confusion, and dementia with disorientation, agitation, or memory loss. Table 2 summarizes the duration of each type of physical and psychological suffering and appendix table 3 lists the results from the literature review on prevalence of the most commonly reported type of physical suffering among patients with serious, complex, or life0limiting health problem. [Table 2: The estimates of duration of each type of physical and psychological suffering by condition]

Most published data on symptom prevalence comes from high or upper-middle income countries where both disease-modifying and palliative treatments are most accessible. Furthermore, most of the literature either focused on physical and psychological symptoms among a single group of patients (such as cancer), or a single symptom (such as pain) in patients with various conditions. Data, mostly from high income countries, indicates that well over 50% of people who die of or live with malignant neoplasms and AIDS experience pain, and that pain is also common among those who live with heart disease, chronic obstructive pulmonary disease (COPD), renal failure, neurologic disease and dementia.(22, 23) Dyspnea (shortness of breath) is especially common among people who live with COPD and heart failure and only slightly less common among those who live with malignant neoplasms and AIDS.(24) Depressed mood and anxiety are widespread among patients with a variety of advanced life-threatening illnesses including metastatic cancer and trauma.(25, 26) There are fewer studies among patients with most other serious, complex, or life-limiting health problems.

Of note, dementia appears both in the list of conditions (Alzheimer’s disease and other primary dementias) and as a symptom of other conditions (HIV/AIDS, cerebrovascular disease, and other neurologic conditions). The term dementia is therefore used in two ways, and the distinction in use of each instance is required.

Identifying multipliers for each condition

The next step in measuring SHS was to determine the proportion of people with each condition who experience SHS. These are called multipliers. Multipliers are mathematical factors that estimate number of people dying or living with SHS based on different data sources. They reflect different strategies applied in the estimation and are provided separately for decedents and non-decedents. For decedents, the multipliers are always a percentage between 0 and 100%, to be applied to total deaths. For non-decedents, the multipliers take one of the three different forms: 1) a percentage between 0 and 100% to be applied to total number of patients living with the disease; 2) a ratio that can go over 100% to be applied to total deaths; or 3) a ratio that can go over 100% to be applied to total decedents in need of palliative care. See table 4 with more details.

To identify the proportion of people with each condition who experience SHS for the different conditions and sub-conditions and therefore identify appropriate multiplies to use for each, an extensive literature review was conducted for both decedents and non-decedents. Empirical evidence of symptom burden for some conditions was identified, but most studies were conducted in high-income settings. Evidence identified from the literature could not directly be used as multipliers since much of it was focused on patients in a certain stage of care whilst the SHS calculation requires multipliers for both people who die within that year – decedents- and another for people who live with a condition – non-decedents. As a result, empirical evidence on percentage of patients with each condition experiencing SHS from the literature review were summarized and presented as the basis of discussion in various expert consensus building exercises. When estimating the SHS burden of non-decedents, experts were asked to consider the SHS burden of an “average” patient for each condition among all patients living with that condition who are not in their last year of life.

Because SHS 2.0 incorporates analysis across a number of years, it was possible to implement improvements to the multipliers for HIV and tuberculosis (TB). SHS stemming from HIV among non-decedents was differentiated between individuals undergoing anti-retroviral treatment (ART) from those who are not, reflecting how the advent of ART and increased access to such treatment revolutionized care for PLWHIV and in turn, SHS associated with HIV. Furthermore, extensively drug-resistant TB (XDR-TB) is differentiated from multidrug-resistant TB (MDR-TB), because antimicrobial resistance and the rise of XDR TB pose major challenges to treatment of tuberculosis which is different from MDR-TB.

For cancer, SHS2.0 also incorporates data across additional years for the estimation of multipliers. In SHS2.0, unlike for the Commission report, five-year survival data were used to estimate non-decedent SHS for malignant neoplasms and leukemia. The GBD data reports only overall survival and does not further disaggregate by years since diagnosis. Hence, the GBD data were adjusted based on the prevalence and mortality data extracted from the Global Cancer Observatory (GLOBOCAN) 2018 (See Panel 1 and Table 3) that report cancer survivorship for 1,3 and 5 years from diagnosis. [Table 3. Multipliers of cancer survivors at 1, 2, 3, 4, and 5 years of diagnosis] A country-specific linearly interpolated trend was applied to estimate prevalence for year 2 and 4 post diagnosis. The approximation of survival was estimated as the ratio between the total deaths and the prevalence in the same period. Lastly, to estimate non-decedent burden for 1990, 2000 and 2010 given that information on 5-year prevalence and survival is not available by year since diagnosis, the GLOBOCAN 2018 data are adjusted using country-income specific quintile distribution data on percentages of all cancer survivors being with each year of diagnosis. (see Table 4 for detail). [Table 4. Percentiles used to impute number cancer survivors at 1, 2, 3, 4, and 5 years of diagnosis in historical years]

Cerebrovascular diseases constitute a major component of overall SHS, yet its non-decedent category was a limitation in SHS1.0. For SHS2.0, non-decedent SHS was calculated for patients living in the year prior to their last year of life, assuming that most patients who live for extended periods with this condition do not experience SHS (as the condition is largely asymptomatic until it becomes serious enough to result in death). Still, data are scarce on the proportion of cerebrovascular disease patients in the final years of life and hence with SHS. An estimate of the proportion of patients who are diagnosed and die in the same year was developed based on a literature search focusing on differences by country income level and this was applied to two years of cerebrovascular disease mortality. (see Tables 5, 6 and appendix table 4). [Table 5: Estimation model used in calculation of cerebrovascular disease patients living with SHS – part 1] [Table 6: Estimation model used in calculation of cerebrovascular disease patients living with SHS – part 2] Because data were not available on the number of deaths per year of patients diagnosed in the last year, a literature search was carried out on the survival of these patients in countries by income level. In other words, the calculation factored in the percentage of newly diagnosed patients that would die within one year as the percentage among all deaths that would occur due to newly diagnosed patients. As literature covering all income groups was not available in all years of interest, i.e. 1990, 2000, 2010 and 2019, missing years and income groups were imputed to the nearest income group and/or to all the year (Tables 8 and 9). The new method limited the estimation of SHS only to patients within the last 1-2 years of their life, since most patients living with cerebrovascular disease can spend years living without SHS. While this method gives us a more realistic estimate of the suffering endured by cerebrovascular disease patients, there is little literature to report an estimate of the percentage of total cerebrovascular disease patients who are in the last 1-2 years of their life (Table 10). Therefore, we applied a series of assumptions plus a limited compilation of data from our literature review to construct the matrix of percentages of cerebrovascular disease patients living within the last 1-2 years of their life by income group, to 1990, 2000, 2010, and 2019. These assumptions are limitations of this study, given the varying strength of the underlying data.

Table 7 presents the multipliers used to calculate SHS for all 21 conditions, separating decedents and non-decedents. [Table 7. Multipliers used to calculate SHS burden for 21 conditions]

Data limitations

First, although a literature review was conducted by condition and symptoms, due to a dearth of reliable empirical data on the types, prevalence, and duration of suffering caused by each SHS associated health condition, both SHS1.0 and 2.0 rely heavily on expert opinion. Moreover, research on palliative care has so far concentrated on Europe and the United States accounting for over 90% of all publications on palliative care but only 15% of the global population. The fact that 85% of the global population produced only 6.5% research publications points to the glaring lack of information on the elements of suffering for the majority of people in the world.(27)

Further, the expert groups are relatively small reflecting limitations in available funding to develop the field of SHS. This makes it especially difficult to develop either disease, region, or country income-specific estimates. The reliance on identifying an “average” patient limits the possibility of exploring regional, cultural or other differences, as well as the effect of providing differential levels of palliative care. The next step in the SHS work is to undertake disease-specific expert panels to refine estimates of people with SHS and especially symptoms and symptom days. This is the focus of research planned for SHS 3.0 and has been piloted for breast cancer and will soon commence on HIV.

Second, there are conditions which generate SHS but are not included in the analysis to-date due to limited scope. For example, chronic paranoid schizophrenia and other severe chronic psychiatric disorders generate severe suffering but are not included in the methods presented here. Another important example is people living in the context of humanitarian crisis,(28) including armed conflict(29) but also climate emergencies, communicable disease outbreaks or those under threat of political, sexual, or ethnic violence who suffer from various types of physical and psychological suffering.

Similarly, our work to date extends to 2019. Estimating the shorter-term SHS that was associated with the coronavirus disease 2019 (COVID-19) pandemic, and the longer-term sequelae for those who suffered the disease should be a key next step in the analysis. This should include the suffering associated with bereavement and the lack of access to palliative care support for caregivers, family members and the community during COVID-19 lockdowns. The wealth of data and publications on the pandemic will make this analysis more feasible.

Family caregivers who experience various kinds of physical, psychological, social, and spiritual suffering as a result of their care work are not included in the estimates. While methods to estimate the types, prevalence, or duration of physical, psychological, social, or spiritual suffering of the main family or informal caregiver have not been within the scope of SHS calculations to-date, this is an important area of future SHS methodological development. Family caregivers typically provide many hours of daily care to patients with serious, chronic, complex, or life-limiting health problems and in many health care settings, especially in LMICs, where they must remain with the patient when admitted to the hospital. Across the world, caregiving work at home and in the communities is predominantly provided by women, and often uncompensated or undercompensated.(30) It has been shown that caregiving can itself represent a source of suffering.(31, 32) Family caregivers may have their own need for palliative care and support in managing bereavement.

Expert opinion provides important information, but a patient-centered approach needs to be included in future work on SHS. Confirmatory research on symptom prevalence and severity with patient- and caregiver-reported real-life data must complement future work. This limitation applies to the symptoms as well as many dimensions of suffering that are important for patients, caregivers, and practitioners. The expert panel identified 11 physical and 4 psychological symptoms, but this is far from an exhaustive list of all possible physical and psychological symptoms patients can experience. Social or spiritual suffering is also not estimated despite being a source of grave concern due to the impact on overall quality of life.(33, 34) In the context of paucity of resources, of poorly organized healthcare systems and of marginalization of large chunks of the population, the impact on the burden of suffering is likely to be considerable.

Further, the quantity of suffering is estimated only in terms of number of people who died from or lived with SHS (SHS 1.0 and SHS 2.0), or the number of symptom days they each experience (SHS 1.0). This approach neglects the intensity or tolerability of suffering experienced. In SHS 3.0, opportunities for understanding the scope and intensity of social and spiritual suffering for patients in need of palliative care will be explored. Gathering patient- and caregiver reported data is the optimal solution to fill in these gaps and should be a priority for donors and foundations interested in improving access to palliative care and achieving the Sustainable Development Goals (SDGs). To date, only a few pilot and exploratory surveys have been undertaken.(35, 36)

Another important area for future work is to determine to what extent suffering can be alleviated with existing practices and techniques at various resource-levels. This also means that the multipliers – percentage of deaths or survivors in need of palliative care by condition – are time-period specific and should change over time based on previously noted endogenous variables, including the change in disease trajectories and their suffering patterns, as health care technologies and systems evolve.

Last but not least, our work looks at one side of the issue: the demand side. It is equally important, if not more, to measure how much of the need for palliative care is fulfilled, by whom, in what quality, and where. Combined with analysis of the actual provision of palliative care, we will be able to identify gaps and provide more tailored policy recommendations.

Future Iterations

The methods described in this paper are pioneering in the field. However, our exploration has only expanded our vision of the bigger, unknown world, leaving more gaps to be filled with future research. Even the more detailed estimate of “symptom-days” – as opposed to number of people – has limitations as a measure of the burden of SHS experienced by patients in the absence of a method to weigh the tolerability or intensity of each symptom. Specifically, the number of days is calculated for each symptom using the available information on symptom prevalences and duration for each condition. Simple aggregation of days with each symptom may lead to overestimation from double counting, as many patients with advanced disease will suffer from more than one symptom at the same time. As such, the Commission report presented two aggregate indicators to evaluate the total symptom burden: 1) the “at least” SHS-days, which equals the symptom-days from the single most prevalent symptom, in most case, pain, of each condition, and 2) the total symptom days should is the sum total of all symptoms. The actual days of suffering experienced by people with SHS should be a number between these two bounds. Ongoing refinement of the calculation of the number of days of SHS experienced by the population in a given year is a core area for SHS 3.0. Moreover, and as described, it is important to note that the calculation of the number of days of SHS is derived from the calculation of the number of people with SHS, not the other way around. As a contribution to measurement of burden, several “summary indicators” or ways to characterize the suffering experienced by patients were developed. Panel 2 presents these secondary indicators that were constructed for the Lancet Commission report. Another dimension that has not been measured to date is to match SHS to an estimate of palliative care need assessment such as the estimated number of required “palliative care visit-days” – the number of days in which a palliative care provider should see the patient, family or caregiver. Symptom days measures only the days during which the symptom(s) persist(s), regardless of whether a visit by or with a palliative care provider is needed. Severe, refractory, or poorly tolerated symptoms may require daily visits while well-controlled symptoms may require a visit only every 2 to 4 weeks. Indeed, provision of effective palliative care can, and should, reduce the number of symptom days as well as the severity of the symptoms. In doing this, palliative care reduces the SHS burden. This remains an area for future discussion and analysis.