Defining Long-Term-Care Need Levels for Older Adults: Towards a Standardized European Classification

ABSTRACT

International comparisons of long-term care (LTC) are hampered by inconsistencies in how to define the need for care. This is especially relevant for the European Union, whose Aging Working Group, which is tasked to project aging expenditure in the long term, has over time used two competing definitions – one based on inability to perform Activities of Daily Living (ADL) and another based on the more subjective Global Activity Limitation Index (GALI). The inconsistency in measurement, as well as problems in defining the intensity of needs, will acquire growing significance as longevity progresses. This paper investigates how the two measures are linked, by analyzing a large European sample survey where respondents replied to both questions. This allows a calibration of the two measures and an investigation of their areas of overlap and difference. The paper concludes by proposing a simple new 4-scale measure of care needs which, by combining the two metrics, introduces some gradation of the intensity of care. Using a consistent measure incorporating intensity, such as the one proposed, will facilitate international comparisons, improve long-term expenditure projections, and aid policy discussion, including the transfer of best practice.

KEYWORDS:

• Long-term care
• public policies for older adults
• Aging
• care need levels
• standardized European classification
• Long-term fiscal planning

Introduction

As Europe’s population is becoming older, the demand for long-term care services will grow (Kaschowitz & Brandt, 2017). Meeting this demand adequately is important for individuals’ well-being but also for fiscal balance. Europe is attempting to rise to this challenge while still lacking a solid working definition of who is in need of care and for how much.

Policy makers, service practitioners, and social scientists use different, and often uncalibrated, definitions of who needs care. Inconsistency can give rise to confusion but is also a source of dysfunctions: Measures of public expenditure on long-term care (LTC) per head are currently spread over a wide range across the EU. This could reflect different needs, cost structures or even institutional arrangements; it could also simply be due to different definitions. In a field where there is an active search for more efficient modes of LTC provision, disagreement on definitions hampers the process of cross-fertilization and the use of best (or worst) practice to deliver innovation. To engage in policy dialogue, common or comparable metrics are important tools. In the case of existing measures in common use, calibration is an important task.

Existing European sample surveys use two different definitions to discern who needs care. Some use Activities of Daily Living (ADL), which measures inability to perform specific activities, while others employ the Global Activity Limitations Index (GALI), a more subjective evaluation of whether a person is hampered in performing “usual” – though unspecified – activities. More strictly, the ADL scale measures if a person has difficulty in performing each of six basic activities of daily living: “eating, bathing, dressing, getting in and out of bed, going to the toilet and walking across a room”; GALI asks if a person is “severely limited” in (unspecified) “activities that people usually do.” The principal mechanism tasked to assist European Union (EU) members to orient their long-term fiscal planning, the Aging Working Group (AWG) has vacillated between the two measures.

This paper addresses some conceptual, measurement, and calibration issues. It does so by focusing on the two measures used by the AWG. It proposes a new indicator which, by combining both, can readily be computed from existing data for all EU countries. It can thus be used as part of international exercises in fiscal planning or in ongoing policy debate. Improving international comparability will aid comparisons and should enable a more fruitful process of transfer of good practice – both exercises of critical importance given the toll the pandemic exacted on populations needing care.

The utility of the proposed new indicator can be illustrated in how LTC is handled in the European Union’s Aging Working Group. The AWG brings together representatives from all member states and advises the Economic and Finance Council in charting fiscal planning to respond to aging populations (Copeland, 2012; Tinios, 2012). In order to measure and project public expenditures on long-term care, it originally used a measure based on Activities of Daily Living (ADL), then switched to a different indicator (Global Activity Limitations Index GALI), causing results to diverge considerably. This switch may have been justified by issues of data availability across all 27 Member States. However, future AWG reports can use a data source, the Survey of Health, Aging and Retirement in Europe (SHARE; www.share-project.org), which since 2017 contains both ADL and GALI for all EU27. Therefore, future reports of the AWG could choose to employ ADL, GALI or a combination of both. This paper takes on the task of, firstly, comparing and calibrating the two measures and, secondly, combining the two in a single ordinal metric. This metric enriches prevalence of need with aspects of intensity and can be shown to be an improvement compared to using either measure in isolation.

The paper starts by examining some conceptual issues: A measure of long-term care need must come to grips with what is being assessed. This is used to present the two alternative measurement scales – ADL and GALI. A discussion of their provenance and meaning is followed by charting how they were used for comparative analysis of their use in the AWG. This sets the scene for the empirical part of the paper – investigating the correspondence between the two measures and calibrating their use using SHARE data for 2015. The paper concludes by examining a new ordinal metric.

Concepts and measures

To compare alternate measures of the need for long-term care, we must be clear about what we are trying to assess. According to the AWG (2012; 2015; 2018): “Long-term care is usually defined as a set of services required by persons with a reduced degree of functional capacity (whether physical or cognitive) and who, as a consequence of this, are dependent for an extended period of time on help with basic and/or instrumental Activities of Daily Living (ADL and/or IADL).” The need for LTC is thus experienced by an individual and must be capable of translating into a demand for LTC services. It follows that, unlike health care, LTC need does not have to be ascribed to a medical or cognitive condition; the same condition could give rise to different care needs. What is involved is the extent of being able to look after oneself, an issue linked to personal independence or self-sufficiency.

The AWG uses the estimated numbers of dependent elderly people per age and gender bracket as key variables to project LTC public expenditures in the long run. However, the numbers needing home-based care that have been used in its various reports on LTC projections resulted from employing two different definitions of need; to these were added administrative estimates of the population in (public) residential care, supplied by each member state. The 2009 Aging Report made use of both: i) in 12 countries, care-need was defined as “facing difficulty with at least 1 ADL,” based on the SHARE survey and ii) in the remaining fifteen countries, care-need was defined as “being severely limited in activities that people usually do” (GALI), based on the EU Statistics on Income and Living Conditions survey (EU SILC) (European Commission, Directorate-General for Economic and Financial Affairs, 2009). Total public LTC expenditures in the base year per age group were estimated by using information supplied by the member states, derived from administrative data, such as from the ESSPROS system (EC, 2009; 2012; 2015; 2018), encompassing various modes of (public) service delivery. Base year expenditures per head were projected forward, using population projections for each country provided by Eurostat; the estimates thus, in principle, include institutional care – at least to the extent that it is reflected in the base year. The report concluded that average public long-term care expenditure for 2060 would lie between 2.3% and 2.5% of GDP by 2060, depending on the demographic or other scenarios used. In contrast, the 2012, 2015, and 2018 Aging Reports projections were based solely on the GALI from EU SILC survey (EC, 2009; 2012; 2015; 2018). The 2018 Aging Report projected public long-term care expenditures in a range between 2.6% and 4.0% of GDP in 2060, much higher than earlier estimates. Such a rise in projected values could reflect a deterioration in the underlying drivers; it could also be due to a change in definitions. The vacillation in the use of sources and definitions could be justified by the availability of data: the only source that covered all member states was the EU SILC survey, which used only the GALI measurement. However, from 2017 data availability is no longer an issue. Starting in wave 7 (2017) SHARE, which contains both ADL and GALI, covers all 27 member states.

Researchers usually utilize the number of ADLs and IADLs for which a person faces difficulties as a threshold for considering this person as in need of care. Some define the threshold as 1 ADL (Ahrenfeldt et al., 2018; Lindholm Eriksen et al., 2013), others raise it to 2 ADLs (Srakar et al., 2015). Others use “1 ADL plus 1 IADL” or even other combinations of ADLs and IADLs (Laferrère & van den Bosch, 2015; Zingmark et al., 2019).

To understand the differences, advantages, and limitations of the two measures, a short background description is useful. The ADL scale was introduced in the United States by Katz et al. (1963), to measure dependency in the case of six activities of daily living: “eating, bathing, dressing, continence, going to the toilet and mobility in the dwelling.” The six were not equivalent: those needing assistance in bathing or dressing could receive it at isolated times in the day, while those in need of eating or toileting needed assistance more frequently. The original scale is hierarchical – that is, someone who cannot eat without assistance will probably need help with the other needs (Katz et al., 1963). Lawton and Brody (1969) subsequently developed the Physical Self-Maintenance Scale by adding a five-degree rating scale of dependency level to the Katz scale; they also introduced Instrumental Activities of Daily Living (IADLs), namely telephoning, shopping, preparing food, housekeeping, doing the laundry, leaving the house independently and accessing transportation services, managing their own medications, handling finances. ADL and IADL data (though not the intensity of need) are regularly collected in the US, among others, by the Health Retirement Survey (HRS), an interdisciplinary panel survey of individuals aged 50+ first conducted in 1993.

The ADL approach was transplanted from the US to Europe in 2004 in the European analog of the HRS, the Survey of Health Aging and Retirement in Europe (SHARE). The same six activities are outlined; however, whereas Katz et al. (1963) used the word “dependent,” SHARE follows the HRS in preferring a vague term, by asking if the individual is experiencing “any difficulty because of a physical, mental, emotional or memory problem,” eschewing any gradation of dependency levels. It is also to note that this wording does not automatically signal a need for care, a general point noted by Cambois et al. (2016).

The GALI was introduced in 2005 in both the EU SILC and SHARE European Surveys (Bogaert et al., 2018). It asks whether respondents “have been limited, and to what extent, in activities people usually do because of a health problem.” By not naming activities explicitly, GALI probably encompasses more activities than either the ADL or IADL measures and introduces a greater element of respondent subjectivity (Berger et al., 2015). These limitations are repeatedly acknowledged by the AWG by noting that survey data using GALI may underestimate some limitations while being too inclusive of relatively minor functional difficulties not requiring care (EC, 2009; 2012; 2015; 2018).

The principal limitation of the ADL scale, as transferred from the HRS and used in Europe, is that it does not incorporate intensity of need, in the way that, say, the Minimum Data Set does in the US (Centers for Medicare and Medicaid Services, 2020). If, for example, a respondent notes difficulties in ADL, but not severe limitations with usual activities, we can reasonably assume that ADL difficulties are not severe and that this person may not actually take up care if that is offered. These nuances are lost when researchers use the ADL scale as the sole criterion of need for care; ignoring the intensity of the difficulty could exaggerate expenditures in projections. In addition, ADL does not capture difficulties encountered in activities not included in the list. Conversely, noting “severe limitations with usual activities (GALI measure),” without at the same time reporting any difficulties with basic or instrumental activities of daily living, could imply that the activities alluded to may be usual but may not be critical for living autonomously. Such may be items not named explicitly in ADL and IADL, for example, driving or using a TV control. In consequence, when the GALI is used as the sole criterion for need of care, as is the case in the AWG, projections may well be affected upward. Thus, both measures on their own have weaknesses which could possibly be overcome if they are used jointly.

Policy makers across Europe, in determining when an individual is entitled to care, try to use more factors in ascertaining the need for care. They invariably start with ADLs, but often complement them with some notion of intensity. All nine European countries studied by Brugiavini et al. (2017) use variants of ADLs and IADLs to ascertain the need for care. Austria uses the number of ADL and IADL limitations. Belgium uses an 8-item ADL scale (6 on functioning, 2 on cognition). The threshold in the Czech Republic is a single ADL. Germany identifies four dependency levels depending on limitations and estimated frequency (Bocquaire, 2016). Italian practice varies by region, though ADLs are still used as triggers. Spain distinguishes three dependency levels. France uses its own scale to yield a dependency index ranging from 1 to 6 (Bettio & Veraschagina, 2010; Joel et al., 2010). Greece employs an ad hoc approach, relying on the judgment of care professionals (Tinios, 2017). Intensity of need is explicitly noted in Belgium and less formally so in Germany.

Τhe AWG projects public LTC expenditure; in consequence, institutional arrangements must be factored in. The number of beneficiaries who are eligible for public long-term care provision affects public economics and the type of care provided (Colombo et al., 2011; Fernandez et al., 2009). According to Bettio and Veraschagina (2010), home care in Europe is delivered through four distinctive modes of provision: (i) comprehensive, publicly subsidized, and administered home care packages, typified by Sweden, (ii) employment of live-in untrained and mostly foreign workers, typified by Italy, (iii) voucher schemes (cheque) exemplified by France, and (iv) predominant reliance on family carers, as in Poland. Each of these types can be considered broadly representative of a larger group of countries. Thus, the same need for home care can give rise to a visit from a trained public employee, an employee of a private firm, or even help from a relative or neighbor, who may or may not receive a public subsidy. The need may also remain unaddressed – giving rise to a “care gap” (Tinios et al., 2022).

Despite heterogeneity in definitions and differences in institutional arrangements, there is much common ground. Once the need for public care is determined, its nature and frequency may vary with intensity, creating variation both between and within countries. Other considerations can include age or severe illness, while different weights may apply to the factors considered in computing vulnerability levels (Brugiavini et al., 2017).

In order to make best use of survey information in policymaking, the existence of an operational and standardized classification of long-term care need and its intensity is important. Consistent concepts could allow researchers, social scientists, and policymakers to work together. The rest of the paper makes a step in that direction, by investigating how ADL and GALI are related; it does so by exploring a survey, SHARE, where each respondent replied to both sets of questions. The quantitative analysis proceeds in three steps: First, we examine whether GALI overstates needs; if true, this would mean that the AWG projections are overestimates. Second, we examine how the number of ADLs is distributed in the population, and whether the data, supplemented by GALI, support the existence of a need-for-care threshold. Third, a new unitary measurement scale combining ADL and GALI is defined; the new scale is easy to implement with existing data and probably captures important aspects missed by either scale on its own.

Methods

Data

Data for this study are drawn from the SHARE (wave 6). SHARE is a multidisciplinary cross-national panel survey of micro data on health, long-term care, socio-economic status, and social and family networks of individuals aged 50 years and over. SHARE started in 2004 and was modeled on the US HRS, whereby respondents are followed regularly after initial contact. We use Wave 6 data conducted in 2015, the last wave before SHARE was extended to all EU member states. That wave covered 18 countries in all, 16 EU members (Sweden (SE), Denmark (DK), Germany (DE), Belgium (BE), Luxembourg (LU), France (FR), Austria (AT), Italy (IT), Spain (ES), Greece (GR), Portugal (PT), Czech Republic (CZ), Poland (PL), Slovenia (Sl), Estonia (EE), Croatia (HR)), plus Switzerland (CH), and Israel (IL).

Data are internationally comparable; translation consistency is heavily tested and checked across countries and waves (Malter & Börsch-Supan, 2017). Though it is a household survey, individuals, once surveyed, are followed in old age homes in subsequent waves. That means that for the original 12 SHARE countries 11 cohorts have been followed into care homes.

SHARE has design features adapted for an older population. Interviewers are trained to overcome problems in interviewing older people, while the panel structure allows answers to be checked against responses in previous waves. There is resort to proxy respondents to respond to cognitive issues or difficulties in communication. SHARE is centrally coordinated, emphasis is placed on faithfulness in translations, while data is checked before public release. We examine only persons aged 65 years old and over, excluding persons 50–64 years of age. This focus on aged care was justified by the remit of the AWG to project all public expenditure linked to aging populations, using the conventional cutoff of 65 years of age. Additionally, in younger groups, the need for care is often assessed by professionals in conjunction with employment or social insurance regulations. Therefore, excluding younger groups protects any classification exercise from cross-contamination from categories used in other fields. This gives a total sample size of 39,737 persons, of whom 31,088 are between 65–80 years of age and 8,649 older than 80. All interviewed individuals over 65 had provided responses to both GALI and ADL.

Measures

SHARE measures both the GALI and the number of ADLs/IADLs. The GALI question asked is: “To what extent have you been limited in activities people usually do because of a health problem, during at least the past six months?” Potential answers are as follows: 1. severely limited, 2. limited, but not severely, 3. not limited. This is identical to the EU SILC wording. The EU AWG considers all those answering severely limited as in need of care. The question asked for ADLs and IADLs uses show cards and is: “Please tell me if you have any difficulty with these activities because of a physical, mental, emotional or memory problem. Exclude any difficulties you expect to last less than three months.” The ADL show card names six ADLs: eating, bathing, dressing, getting in and out of bed, going to the toilet, and walking across a room. The IADLs are nine: shopping, cooking, doing laundry, performing housework, taking medication, using the telephone, managing money, leaving the house independently/accessing transportation, and using a map in a strange place.

Not all instrumental problems can be expected to have equal impact on the ability to live autonomously. Some, which we here call “domestic IADLs” (dIADLs) imply an inability to cope without the active presence of a helper; others, dubbed “social IADLs” (sIADLs), may require occasional help outside the house, with a smaller expected impact on autonomous living. dIADLs are as follows: shopping, cooking, doing laundry, performing housework, taking medication, and using the telephone. sIADLs are as follows: managing financial issues, leaving the house independently/accessing transportation, and using a map in a strange place. Our presumption is that sIADL activities imply a lower need for care; if one reports sIADLs solely, a lower need for care is warranted.

Statistical methods

The analysis investigates the correspondence of ADL and GALI as alternative measures of the need for care, applied to the entire sample of SHARE respondents aged 65 or over. It relies heavily on descriptive measures and graphical presentations. The level of statistical significance reported is .05.

The relationship between ADL and GALI is further investigated by a probit model, where the relationship between the GALI statement of “severe limitations” is used as the dependent variable and is related to ADLs and to a range of cofactors. The logic is that GALI (the dependent variable) reflects respondents’ generalized and subjective evaluation, which results from the specific needs itemized as difficulties in the ADLs and IADLs questions, as well as from other unspecified sources. This evaluation is mediated by how the individual interprets those difficulties and judges their intensity. This subjective filter may give valuable systematic information about what a person considers serious enough to warrant care in his/her case. Stoic individuals will likely express a higher personal threshold of inability to help themselves. Additionally, where an official provider applies a similar classification, this may be internalized in the personal evaluation. The way GALI is linked to ADLs and IADLs is likely to vary with individual characteristics (age, gender, existence of serious illness). It may also differ systematically by country. The difference could be due to how a national culture values a stoic disposition; more importantly, it may reflect how the criteria used in the administration of public LTC are internalized by respondents. The influence of age is allowed to be non-linear by including a quadratic term for age.

Finally, the paper proposes a new indicator combining the two measures. Its value added is investigated by examining whether it is associated with receiving professional care.

Results

The results are divided into two sections: (1) calibrating ADLs and GALI as measures of care need; (2) a probit analysis of GALI as a function of ADLs and cofactors.

Calibrating ADLs and GALI as measures of care need

The proportion of persons of aged 65 years and older considered in need of care per definition type is reported in Figure 1. One in 10 (10.3%) is judged as being in need of care with the more stringent definition (at least 2 ADLs). This doubles if the GALI definition is used (20.8%). The proportion falls in between these extremes with the other definitions, at least 1 ADL and 1 dIADL (14.9%), and at least 1 ADL (18.8%). These findings prove beyond doubt that the GALI entails a broader definition of care than the ADL measure, which presumably will also be reflected in any expenditure projections.

Figure 1. Proportion (%) of persons 65+ in need of care per care-need definition with 95% confidence intervals: “at least 1 ADL plus 1 dIADL,” “at least 2 ADL,” “GALI.”

Table 1 probes the correspondence between the two measures further, by reporting the GALI score for each number of ADL. As the number of ADL limitations increases, the severity of limitations with usual activities (GALI) also rises. However, less than half of the persons facing difficulties with exactly 1 ADL (44.0%) report severe limitations with usual activities (86% report some kind of limitation). The presence of a severe limitation rises to 60% in the case of 2 ADLs and more than 74% after 3 ADLs. If the criterion of need is “more than 1 ADL,” 8.3% report not even a partial limitation, presumably judging that they can work around the obstacle named. Conversely, 11.5% of those not reporting ADL judge themselves “severely limited” in (unspecified) usual activities.

Table 1. Correspondence of ADL and GALI, 65+: Number of ADLs by GALI seriousness, entire sample.

		GALI – Seriousness of limitations in usual activities by number of ADL difficulties
		GALI	no GALI*	no GALI*
Number of ADL difficulties	Proportion (%) of persons per number of ADL difficulties	(severely limited)	(limited, but not severely)	(not limited)	Total
(1)	(2)	(3)	(4)	(5)	(6)
0 ADLs	81.2	11.5	35.8	52.7	100.0
1 ADL	8.5	44.0	41.8	14.2	100.0
2 ADLs	3.7	59.8	34.8	5.3	100.0
3 ADLs	2.0	73.5	23.1	3.3	100.0
4 ADLs	1.4	78.6	19.4	2.0	100.0
5 ADLs	1.3	91.1	6.9	1.9	100.0
6 ADLs	2.0	91.9	5.6	1.8	100.0
Total sample	100.0	20.9	34.8	44.3	100.0
At least 1 ADL	18.8	61.1	30.6	8.3	100.0
At least 2 ADLs	10.3	75.1	21.4	3.4	100.0

ADL: Activities of Daily Living; GALI: Global Activity Limitation Indicator.

*Note: Not included in the category of persons in need of care in the Aging Working Group approach. Only severely limited are considered persons in need of care.

Source: SHARE wave 6 data n = 39,737.

Table 1 also reports the distribution of ADLs in the 65+ population. Focusing on the fifth of those facing some ADL difficulty, almost half (8.5% of 65+, or 45.1% with ADLs) face difficulties with a single ADL; this concerned difficulties with dressing or bathing (50.1% and 32.7%, respectively), rather than one of the other activities.

Table 2 examines how far the inability to perform ADLs translates to instrumental ADLs, as well as the two kinds of IADLs distinguished. The inability to function (IADL) rises precipitately after 1 ADL, from half of those with 1 ADL to over three quarters, if there are two or more. Moreover, the presence of domestic IADL rises sharply with 2 ADLs and tends to become the rule after 3 ADLs. In contrast, social IADL rises more gradually.

Table 2. Association of ADLs and types of IADLs: of dIADLs and sIADLs by number of ADL difficulties, 65 + .

Number of ADL difficulties	Share (%) of persons with at least 1 dIADL	Share (%) of persons with at least 1 sIADL	Share (%) of persons with both dIADLs and sIADLs
0 ADLs	11.2	9.0	4.9
1 ADL	54.0	41.1	33.0
2 ADLs	79.2	67.0	62.0
3 ADLs	91.7	78.6	76.1
4 ADLs	95.6	87.6	86.4
5 ADLs	98.5	97.8	96.6
6 ADLs	99.9	98.7	98.6

ADL: Activities of Daily Living; IADL: Instrumental Activities of Daily Living; dIADL: domestic Instrumental Activities of Daily Living; sIADL: social Instrumental Activities of Daily Living.

Source: SHARE wave 6 data n = 39,737.

Figure 2 is a Venn diagram, which explores the considerable overlap, but also divergences, between the two measures. Persons encompassed within the shaded area ABCA (7.4% of the population of age 65+) would not be considered in need for care if the GALI definition is used; the same persons would be deemed to need care if the ADL definition is used. The reason is that, despite mentioning an ADL, they are not “severely limited”; the respective percentages involved are noted in the Venn diagram and can be derived from Table 1 (multiplying columns 2 and 3). Those who mention at least two ADLs deem them important (via the GALI reply) and lie in the intersection of the two sets (area FDGF, 7.7% of the population of age 65+); they are 8 out of 10 of those reporting 2 or more ADLs, indicating more acute needs. Finally, we should note a not insignificant number who report a generalized GALI without mentioning an ADL (sector JCHA, 9.4%).

Figure 2. Correspondence of ADLs and GALI: Venn diagram, persons 65 + .

Notes: Proportions refer to proportion of the population of age 65 +

Probit analysis: How do limitations with ADLs correlate with GALI?

Table 3 reports a probit analysis relating the more subjective evaluation (GALI) to the relatively “objective” statements ADLs, IADLs, as well as to cofactors hypothesized to filter this evaluation process. The marginal effects show how each additional ADL, dAIDL, and sIADL increases the probability of declaring severe limitations in usual activities (GALI), controlled for the following covariates: age, age squared, chronic illnesses, gender, and country origin. Each additional ADL increases the probability of facing severe limitations in usual activities (GALI) by 9.9 percentage points for the general population aged 65 years and older, all else being equal. Its effect is highly non-linear, diminishing as the number of ADLs increases. Instrumental ADLs exert an independent influence, though as expected GALI is more sensitive to dIADL than sIADL: each additional difficulty mentioned increases severe limitations by 4.9 percentage points as opposed to 3.3 percentage points. Age has a non-linear effect: its independent impact rises but at a diminishing rate.

Table 3. What causes severity? Probit analysis of GALI by ADL, persons 65+, (Marginal effect of ADLs, dIADLs, and sIADLs on the probability having severe limitations in usual activities; GALI).

	Total sample of persons, age 65+ (n = 39,707)
Υ = prob (GALI: severe limitations in usual activities)	dF/dx	std. err.	p
number of ADL limitations	.099	.009	.000***
Number of ADL limitations squared	−.011	.001	.000***
number of dIADL limitations	.049	.005	.000***
number of sIADL limitations	.033	.007	.000***
Age	.035	.010	.000***
Age squared	−.0002	.0001	.001***
Pseudo R-square	.357

#p < .10, *p < .05, **p < .01, ***p < .001.

Models adjusted for covariates: gender, chronic illness, country origin.

ADL: Activities of Daily Living; dIADL: domestic Instrumental Activities of Daily Living; sIADL: social Instrumental Activities of Daily Living; GALI: Global Activity Limitation Indicator.

Source: SHARE wave 6.

Discussion

Our findings confirm the supposition of Berger et al. (2015) and EC (2018) that the GALI involves a broader definition of need for care than both the ADL and IADL measures considered. In this respect, a decision (as by the Aging Working Group) to pick GALI to project LTC needs would tend to overestimate base-year needs. At the same time, it excludes those who report difficulties with one or two ADLs but do not report severe limitations in usual activities. These differences will have a direct impact on LTC public expenditure projections but could also affect policy design.

Combining ADL with GALI indicates that facing difficulties with two ADLs presents a kind of threshold or point of inflexion for the need of care: the vast majority of those facing difficulties with three or more ADLs report severe limitations with usual activities, while for those with two ADLs, the respective proportion reporting severe limitations is less than 60%. Another trigger is facing difficulties with one ADL, but only if combined with a dIADL. A reason could be that most of those persons facing difficulties with only one ADL, face difficulties with bathing or dressing; these two are considered less demanding compared to the other four ADLs (Katz et al., 1963). Thus, when at least one domestic IADL (though not social) coexists with one ADL, it appears to trigger a declaration of severe limitations. These findings echo the existing literature, where the most common thresholds used to signal care need are facing difficulties “with 1 ADL” or “with 1 ADL plus 1 dIADL” or “with 2 ADLs.”

These thresholds, when combined with the GALI, could be used to define a new metric of the need for care, which enriches ADLs with a notion of intensity. Employing a more granular care need indicator computable from existing sources could facilitate international comparisons of different modes of service delivery provision, of the criteria used, and help the cooperation of public and family provision. (Brugiavini et al., 2017; Colombo et al., 2011; EC, 2018; Fernandez et al., 2009). A structured policy dialog could encourage the spread of best practice, enhance cooperation, and lead to gains in policy coordination – which after all is the objective of the Open Method of Coordination.

Based on the discussion above, a 4-point scale that categorizes care level needs is proposed. This is defined in Table 4 and illustrated by means of a Venn diagram in Figure 3. The new measure draws information from three categorizations: ADLs (the elongated horizontal ellipse in the Venn diagram), GALI (the ellipse pointing right) and the two types of IADLs (the vertical and left-pointing ellipses). The combination of these criteria can be used to yield the four-level tiered measure outlined in Table 4: minimum, low, moderate, and high. The relevant sectors are marked in the Venn diagram with darker shades implying higher need.

Figure 3. The four-scale needs scale: Venn diagram.

Table 4. Proposed Ordinal Classification of Long-Term Care Need Levels (LTCNL).

0	No need for care	without any ADL, dIADL, sIADL or GALI limitations
1	Minimum need for care	no ADL, no dIADL but at least 1 sIADL (regardless GALI) or no ADL, no dIADL, no sIADL but severe limitations in usual activities (GALI)
2	Low need for care	1 ADL without any dIADL or no ADL but at least 1 dIADL (regardless GALI)
3	Moderate need for care	2 ΑDLs without “severe limitations in usual activities” (GALI) or 1 ADL plus at least 1 dIADL (regardless GALI)
4	High need for care	3 or more ADLs (regardless GALI) or 2 ΑDLs with “severe limitations in usual activities” (GALI)

ADL: Activities of Daily Living; IADL: Instrumental Activities of Daily Living; dIADL: domestic Instrumental Activities of Daily Living; sIADL: social Instrumental Activities of Daily Living; GALI: Global Activity Limitation Indicator.

Using the SHARE data, Figure 4 categorizes the proportion of people aged 65 years and older who need care according to the proposed metric. Using this for the entire 65+ sample, 8.9% of persons would be deemed to be in high need of care, 6.1% in moderate, 13% in low and 8.6% in minimum need. The category high need of care is narrower than the measurement of at least 2 ADLs. At the other end, being considered at any level of need for care is looser than the GALI measurement. In that way, we avoid the over-simplification of a binary definition, and potentially enrich policy discussion. Combining the two metrics could enable researchers, while still making use of existing data sources, to move toward a simple pan-European four-scale ordinal classification of long-term care needs with more comparable thresholds. These thresholds could rationalize comparisons, notwithstanding existing data limitations in Europe.

Figure 4. Proportion (%) of persons 65+ in need of care by level of seriousness, per country.

Would the proposed indicator add value to the study of LTC and aid international comparisons? A possible test is whether need as ascertained (and graded) by the person concerned can predict reactions to need better than the other available indicators. For example, for a given system, greater need could be inferred if the new indicator exhibits a higher correlation with the provision of care or, better still, with a measure of volume, care hours or expenditure. In the absence of an independent dataset, we can use SHARE to compare the three measures: ADL, GALI, and new combined LTCNL. The proxy for the system response we use is whether an individual receives professional care – which in this case encompasses both public and private bought-in services.

Table 5 contrasts four attempts to predict the probability of receiving professional care. It would be reasonable to assume that resorting to professional care serves as an indirect acknowledgment that the need is greater. Although some individuals are limited to professional care by necessity (non-availability of family caregivers), while others may prefer it, we would expect to see the incidence of professional care to act as a (possibly noisy) indicator of greater subjective assessments of need. The benchmark is a model which uses only “objective” information of the kind known to service providers – cofactors such as age, gender, long-term illness, and country; these are independent of any specific needs assessment. This “blind” model is contrasted with models adding some measure of expressed need – GALI, ADL, and the LTCNL. We know that, in our sample, those indicators are highly correlated. Rather than a rigorous statistical test (which would contrast each indicator with the others), we can use a weaker criterion – the proportion of the observed variance “explained” by the model (the “Pseudo R-squared,” which generalizes the familiar indicator in the case of discrete dependent variables; Wooldridge, 2002, p. 465). A more operational criterion is the extent of the needs indicator “absorbs” explanatory power from “objective” information such as age.

Table 5. Probability of receiving professional care – Alternative models using different types of need indicator, Probit model, and selected indicators.

		Coefficient of		Needs indicator
	Pseudo R^2	Age* E^2	(age)^2* E^2	Minimum	Maximum
Only cofactors	0.1850	−1.38*	0.02***	-
Cofactors + GALI	0.2170	−1.12	0.01**	0.155***
Cofactors + >1 ADL	0.2239	−0.37	0.01#	0.176***
Cofactors + LTCNL 4 scale	0.2627	0.05	0.00	0.099***	0.407***

#p < .10, *p < .05, **p < .01, ***p < .001;

Models also adjusted for: country origin, gender, long-term illness existence.

Note: Respondents living in care homes (1.5% of the sample) excluded.

Source: SHARE wave 6▐

Table 5 proceeds to such a comparison using the same dataset as before. It starts from a simple model containing only cofactors and contrasts that with one where, alternatively, GALI, ADL, and LTCNL are added to the original cofactors. We see that the explanatory power (in the sense of correctly predicting the receipt of care) rises considerably when the new indicator is employed (from 0.185 to 0.263). Perhaps more importantly, once LTCNL is known, age ceases to matter completely. This impact is more marked with LTCNL than with either GALI or ADL. These statistical findings illustrate that, once we take on board individuals’ own subjective gradations of need, our ability to “predict” the resort to professional care improves dramatically. The statistics also indicate that age appears only to stand in as a proxy for need; once independent estimates of needs are introduced, it ceases to matter.

Limitations

The SHARE is a household panel survey, which excluded persons living in nursing homes from the initial sampling frame; however, original respondents who moved to care homes were not dropped. In contrast, its main alternative, the EU-SILC sampling frame consists exclusively of households. Thus, though care homes are not excluded in SHARE, they are likely to be underrepresented. Moreover, attrition from the panel sample is likely to be higher for those transferring residence and more so for those moving to nursing homes. Results for countries which have been on the SHARE panel the longest or which rely on nursing homes are not totally unrepresentative of the situation on the ground (11% of those with at least 1 ADL live in a care home in the Nordic countries, 5% in Central Europe, but only 2.2% in Southern Europe and 0.9% in the East). Conversely, the SHARE sample reflects more closely that part of the older population living in the community. This group accounts for most community-based LTC, and is, arguably, of greater importance in planning for future LTC needs, whether public or private.

We know that much care is concentrated in the last 6 months of life. These are the subject of a separate SHARE “exit interviews” completed by relatives after the death of a member of the panel. These were not included in our analysis; similarly to the previous point, this is unlikely to be a major factor for community-based LTC.

Sampling of the very old and frail, who are also most heavily in need of LTC, is fraught with difficulties. As a result, this group is often underrepresented in surveys of the general population. SHARE, being a survey geared toward the study of aging, guards against that, through surveying all individual residents in a household, the use of proxy respondents, and keeping track of non-response biases (Malter & Börsch-Supan, 2017). However, it is indubitable that difficulties will most likely persist.

A concern with international surveys is the fidelity and consistency of translations. The SHARE questionnaire is translated into 18 languages. However, special attention is placed on consistency in translations; moreover, the repetition of the survey and its panel nature should guard against any major issues. A more important concern is the impact of different institutional structures, which exist for the provision of services. In countries relying on formal systems of provision, which invariably involve judgments of severity, the bureaucratic classification will most likely influence individuals’ own perceptions. In more familial systems, such as those in the South and East of Europe, such an influence will be more limited. Subjective evaluations are filtered by answering styles, attitudes to being stoic (“stiff upper lip”) or to health conditions which “justify” needing care; all these might vary from country to country or by gender. Such influences are allowed through the presence of co-factors and country dummies in the probit analysis. The presence of other co-factors or non-linearities cannot, however, be ruled out.

The proposed composite measure is based on ADL and GALI, two measures with differences in the time frame used; GALI asks for the respondent to reflect back over 6 months, whereas for the ADL questions asks the respondent to look forward indefinitely and to not consider short-term (less than 3 months) functional limitations. Although both time frames try to exclude short-term limitations, the difference might have an impact on care-need assessment.

Finally, “severity” as judged by the people affected may not be the same as “severity” in the sense of necessitating more intensive and/or more costly services. Thus, the intensity as judged by the method proposed may not immediately correspond to differences in service inputs or of cost, which will be the items of interest in projection exercises. Using this classification to improve projections would probably entail a further step of analyzing how differences in needs translate to differences in inputs or costs.

Conclusions

As LTC grows and longevity proceeds, the need for an informed policy discussion across Europe grows apace. The overdue nature of this discussion was painfully demonstrated by the toll wrought by Covid-19 on beneficiaries of long-term care throughout the EU; this toll bore little relation to per capita public expenditure and is serving to underline the urgency of exploring modes of provision based on partnership and complementarity between the family and formal systems. Such discussions could benefit from access to a standardized vocabulary and a European categorization of long-term care needs, as well as from awareness of how different measures interact. Policy discussion, in the AWG or elsewhere, could avoid systematic overestimates that come from using GALI and, in this way, improve the definitions for need of care. Classifying the extent of care needs will, thus, be a useful step. The 4-point scale of care-needs proposed combines ADLs, IADLs, and GALI. It is simple and represents a practical improvement that can be implemented immediately. It draws on information currently available (at least in SHARE) and improves on current practice by combining questions from three different, yet linked, domains. It would also allow different stakeholders – public policymakers, researchers, private insurance companies – to focus on, compare, and evaluate the needs of different population groups.

Focusing specifically on EU projections, continued use of GALI as a stand-alone indicator of the need for care entails a serious overestimate of expenditure. It is thus recommended that the AWG discontinues its use. A simple, yet more sophisticated, approach, such as the proposed 4-point needs scale, could be an immediate improvement and would add value to projection exercises used by decision makers. Going beyond projections, central guidelines on thresholds regarding LTC need levels could be discussed as a method to improve analysis and service delivery among EU countries. Policy makers within countries could use them as benchmarks to evaluate and benchmark their national approach. The final result could lead, with time, toward a convergence of European LTC policies.

The new scale could also aid the design of public policies encouraging interoperability and cooperation between public, private, and family service provision. As an example, the public sector could prioritize providing LTC services to those in significant and moderate need, the private sector could focus on moderate and low need of care and the family and individual social networks on persons in low need of care and providing help with social activities.

Key points

• Europe lacks a consistent definition of needs for long-term care (LTC), an issue which, as longevity progresses, will rise in significance.

• The European Union Aging Working Group overestimates LTC needs, by using the “global activity limitations index” (GALI) as the sole care-need indicator.

• Activities of daily living (ADL) and GALI as measures of LTC need are compared and calibrated using a large European sample of people aged 65+.

• A standardized classification of care-need levels is proposed combining GALI, ADLs, and instrumental activities of daily living (IADLs).

• GALI can complement ADLs and IADLs by introducing some element of severity of need.

Acknowledgments

This paper uses data from SHARE Wave 6 (DOI: 10.6103/SHARE.w6.600), see, Malter and Börsch-Supan (2017) and Börsch-Supan et al. (2013) for methodological details. The SHARE data collection has been primarily funded by the European Commission through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812) and FP7 (SHARE-PREP: N°211909, SHARE-LEAP: N°227822, SHARE M4: N°261982). Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the U.S. National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C) and from various national funding sources is gratefully acknowledged (see www.share-project.org).

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This work was supported by the European Commission under the Horizon 2020 Programme (H2020), as part of the project SHARE-COVID19 (grant agreement no. 101015924).