How well does performance predict self-reported functioning in - - PDF document

1 How well does performance predict self-reported functioning in Europe? A cross-country comparison among 14 European countries using SHARE data. Maria Bilo1, Eileen Crimmins2

1 Dipartimento di Scienze Statistiche, Sapienza Università di Roma, Italy 2 Davis School of Gerontology, University of Southern California, USA

Introduction Measurement of physical capability can be based on either performance or self-reports. Both approaches have advantages and disadvantages when measuring functioning among older adults [Angel et al. 2000, Gill 2010]. Self-reported functioning is often used because it is low cost and easy to administer; there is no need for equipment or special training of interviewers. However, respondents may evaluate their functioning relative to their own environment, including their personal adaptations, and relative to their reference population, resulting in measures which are not comparable across members of the population or across populations [Coman & Richardson 2006]. Performance-based measurement of functioning is more expensive given the need for special equipment and training and the time required for

• administration. However, results may depend on the respondent’s motivation and reflect

maximal performance in an artificial environment rather than every day or normal ability. Comparison of the two approaches indicates that they can provide disparate results [Roedersheimer et al. 2016, Sager et al. 1992] because they indicate different aspects of functioning [Bean et al. 2011]. For example, performance-based measures seem to express strength or velocity related functioning better than self-reported measures. On the other hand, self-reports appear to be a better measure functioning that relies on aerobic capacity, disease burden or self-efficacy. Furthermore, Bean et al. point out that performance-based measures primarily reflect abilities of the lower limbs; whereas self-reported measures indicate abilities in tasks that require both the upper and lower limbs. A general conclusion is that both types of

2 measures capture information on the functional ability of older people but in different dimensions [Seidel et al. 2011]. Comparison of functioning across national populations show different results depending

• n the method of measurement. For example, differences in performance based functioning

across Europe indicate that older persons in the Northern-continental countries have higher grip strength than those in Southern European countries [Andersen-Ranberg et al. 2009]. Furthermore, it was shown that grip strength indeed strongly predicts disability, morbidity and death among older Europeans [Hairi et al. 2009] but that socio-economic drivers (wealth, education, occupation and income) of grip strength do not consistently predict grip strength in Europeans. The aim of this paper is to examine cross-national differences in self-reported physical functioning and performance-based functioning. The questions addressed include: 1) Are there statistically significant differences in self-reported functionality among European countries? 2) Do differences in functioning ability as evidenced by performance-based measurement explain the differences in self-reports of functioning differences across European countries? Background As mentioned, there are several studies that compare either performance-based or self-reported measurements between countries with or within Europe. The US, e.g. shows higher self- reported disability than the Netherlands [Kapteyn et al. 2007]. Another study of disability in three European countries illustrated cross-cultural variation in the association of self-reports and performance-based disability measurement [Van den Brink 2003]. Data from 1990 were used to compare men from Italy, the Netherlands and Finland. The initial North-South gradient in self-reported disabilities as well as performance abilities was explained by adjusting by age, socioeconomic status, household composition and chronic diseases. Furthermore, it could be shown that performance score was positively associated with self-reported disability score in

3 those three countries and that the strength of this association did not differ between the

• countries. The study concludes that those differences among European countries may result

from different sociocultural and physically environmental backgrounds of the participants. [Andersen-Ranberg 2009, Brothers et al. 2014, Verropoulou 2009] We expect national levels of both observed and reported functioning ability to be affected by sociodemographic composition of populations. Educational level has been shown to be a strong predictor of functioning [d’Uva et al. 2008, Verropoulou 2009]. Gender has also been shown to be related to functioning ability [Crimmins et al. 2011, Verropoulou 2009]. It has also been shown that performance-based tests explain most variance in self-reported health, even though there was no strong association between self-reported and performance-based health [Kempen et al. 1996]. Nevertheless, it is unclear how and to what extend those performance-based measures influence self-reported health in different countries. Comparative studies are essential to understand circumstances that are associated with health differences among different countries. To understand how certain factors have a possibly different impact among countries provides a great potential to expand the knowledge around health differences and the approaches to minimize them. Comparative studies improve the understanding of the limits of the generalization of health issues in different countries. Consequently, we reckon that performance-based test results play a part in contributing to predict self-reported health but that those influences vary by countries. Methods Data The data used is drawn from the Survey of Health, Ageing and Retirement in Europe (SHARE) [Börsch-Supan 2016]. SHARE is a cross-national panel survey producing individual level data

• n health, socio-economic status, and social and family networks for a number of European

countries. This paper uses data from SHARE Wave 2 (2006/2007) (DOI:

4 10.6103/SHARE.w2.500). All data were collected by interviewers in a face-to-face computer- aided personal interviews (CAPI), supplemented by self-completion paper and pencil questionnaires [Börsch-Supan et al. 2008]. See Börsch-Supan et al. [2013] for methodological

• details. The second wave of SHARE (2006/2007) interviews people aged 50 and older from

14 European countries: Austria, Belgium, Switzerland, Germany, Denmark, Spain, France, Greece, Italy, Netherlands, Sweden, Czech Republic, Ireland and Poland. For our analysis, we select first people aged 50 and older from those countries with reliable information about their age and sex (N = 33,697). Furthermore, we excluded respondents who did not perform at least 1 out of 3 performance-based tests (grip strength, peak flow, chair stand) from our analysis (2,106 cases). SHARE also has a fourth performance-based test of walking speed. This test was not used in our analysis because only respondents age 75 and higher were asked to do this test. Therefore, we start our analysis with 31,591 observations. Measures Self-reported functioning ability is reflected by an additive index, based on responses to questions on 6 activities of daily living (dressing, walking across a room, bathing or showering, eating, getting in or out of bed, using the toilet), 7 instrumental activities of daily living (using a map for orientation, preparing a hot meal, grocery shopping, making telephone calls, taking medications, house/garden work, managing money), and 10 Nagi indicators of mobility, arm function and fine motor limitations (walking 100 meters, sitting for about 2 hours, getting up from a chair after sitting for long periods, climbing several flights of stairs without resting, climbing one flight of stairs without resting, stooping/kneeling/crouching, reaching or extending arms above shoulder level, pulling or pushing large objects like a living rooms chair, lifting or carrying weights over 10 pounds/5 kilos, picking up a small coin from a table). People who respond that they are limited in an activity are scored a 1. Therefore, the ADL variable has a value range from 0 to 6, whereas the range of the IADL variable spans 0 to 7. The NAGI

5 variable has values from 0 to 10. Thus, all three variables contain information about how many activities are in any degree limited for the respondent. Hence, the index of self-reported functioning is the sum of the number of activities the respondent had any troubles with – ranging from light execution problems to not being able to perform the task. Consequently, the value of the index ranges from 0 to 23 and is used as a continuous variable. Functioning ability was also measured using three performance tests: lung function, grip strength, and a chair stand. The peak flow test measures the lung capacity of the participants by exhaling into a peak flow meter, which displays the liters of air breathed out per minute (l/min). By measuring how fast a participant is able to breath out the results can indicate whether the airways are narrowed. The value of this variable ranged from 30 to 999 l/min. Grip strength was measured with the help of a dynamometer handle which measures the exerted force in kg while pressing it. The best two results out of four tries with each hand were recorded. The level

• f grip strength is an important indicator of upper body strength. The value range of this variable

was 1 to 84 kg. For the chair stand test, the participant moves from a sitting to a standing position five times without using his or her arms. This procedure is measured in seconds and tests the lower body strength and endurance of the participants. The values of the variable ranged from 0 to 99 seconds. If a respondent was not able due to medical reasons or concerns to perform a performance-based test, he/she was given the worst value measured in this test. This was true for about 3.12% of participants for the peak flow test, about 1,63% for the grip strength test, and around 11.82% for the chair stand. Additionally, we used the listwise deletion concept [Diekmann 2011] to delete missings from the sample that represented less than 3% of the respective variable. This could be applied to the dependent variable (self-rated health), the education variable as well as the peak flow and grip strength variable. Chair stand showed too many missings (19.5%) and therefore, the missings will be kept. Chair stand will be integrated last in the regression to avoid several

6 fluctuations of observations through the regression process. The N we start our regression analysis with is therefore 30,036. All three performance-based variables are used as continuous measures. The same applies for age and education which are both measured in years. Analysis To examine the relationship between measured and self-reported functioning, we conducted Tobit regressions with self-reported functionality as the dependent variable. To estimate the relationships between the variables, we used Tobit regressions because of the distribution of the dependent variable; 49.4% of the sample do not self-report any limitation in functioning, so the distribution is highly left skewed. Variables are added to the regression model one at a time to note changes in differences across countries with each additional parameter [Hair et. al. 2010]. Using self-reported functioning as the dependent variable, we estimated 5 regression models with the following independent variables: 1) age, sex and nationality; 2) education; 3) grip strength; 4) peak flow and 5) chair stand. Descriptive statistics The sample size and the weighted sample characteristics by country are represented in table 1. The different sample sizes for each country have to be considered when investigating the statistical significance of the results. Around a quarter of the participants are from Germany. Other countries contributing large numbers of participants include Spain, Italy and France (11, 16 and 16%) as well as Poland (~10%). Austria, Sweden, the Netherlands, Denmark, Greece, Switzerland, Belgium, Czech Republic and Ireland each contribute between one and five percent to the survey pool. The average number of self-reported limitations is 1.9, with the highest in Poland (3.3) and the lowest in Switzerland (.9). Mean age of the sample (64.5) is lowest in Poland (62.8)

7 and highest in Austria (65.4). The average proportion of females is quite similar across countries, ranging from 51% in Sweden to 56% in Poland. The mean education of the sample is 10.5 years with the lowest mean of 8.0 in Spain and the highest mean of 13.1 years in

• Denmark. The average result on the peak flow test is 307 l/min with the best mean result in

Sweden (390) and the worst mean result in Italy (263). The by far lowest mean in grip strength is found in Spain with 30.9 kg and the highest mean in Sweden (36.6), whereas the sample average lies at 34.2 kg. The chair stand test results show a wide range as well. While the average test score is 25.9 seconds, the best mean result is 14.5 (Denmark) and the worst is 36.1 seconds (Italy). Results Graph 1 shows the change of all country-specific coefficients from the Tobit regression according to the adjusting variables that are added to the regression from the highest (Poland) to the lowest (Switzerland). An education effect is visible and manifests oneself through an

• verall decrease of the coefficients. An exception in this trend can be seen in Denmark where

the coefficient increases ( = .113) after education was added. The biggest coefficient decreases can be seen in Spain, Italy and Greece ( = - .972, - .896, - .825) and the smallest in Czech Republic (=- .050). If the performance-based measurements are taken into account, the coefficients continue to decrease for around half of the countries (Poland, Italy, Ireland, Spain, Greece, France, Czech Republic) while they increase for the rest (Belgium, Denmark, Austria, Netherlands, Sweden and Switzerland). The highest decreases experience Poland and Italy ( = - 1.155, - 0.902) while Switzerland experiences the by far highest increase ( = .716). Even though Switzerland experiences this increase after performance-based measurements are taken into consideration, it still has the lowest coefficient (-1.498), closely followed by Spain (-1.410). Poland has the by far highest coefficient (1.454), followed by

8 Germany as the reference. The Southern countries seem to have remarkably lower coefficients than the Northern countries after education and performance is taken into the equation. Table 2 displays the first regression model, which contains self-reported functionality as the dependent variable and age, sex (males as reference) and nationality (Germany as reference) as the independent variables. Most of the countries show a significant coefficient but with wide ranges. The lowest significant coefficient is -1.999 (Switzerland) while the highest is found in Poland (3.271). Age shows a highly significant influence on self-reported functionality in the survey population with a coefficient of 0.220. Therefore, it can be concluded that if a participant were to be one year older, his expected self-report of limitations would increase by 0.220. Gender shows a highly significant coefficient of 1.884, which means that female respondents are expected to have a by 1.884 increased self-reported limitation score. The second regression model adds education as an independent variable. Education has a highly significant coefficient of -0.202, which means that if a participant were to have one year more education, his expected self-report of limitations would decrease by 0.202. The coefficients on country of origin decrease overall as already mentioned. The third model adds the grip strength test results as the first performance-based functionality measurement. Grip strength is highly significant and indicates that if a participant were to have a test score with one kilogram higher, his expected self-report of limitations would decrease by 0.161. Country coefficients experience an overall decrease after grip strength is added with the exceptions of Austria, Sweden and Switzerland. The lowest coefficient can still be found in Switzerland (-2.04) and the highest in Poland (2.096). The fourth model adds the peak flow test as a variable to the regression. The variable for the peak flow test is highly significant as the other performance-based variables are. It indicates that if a participant were to have a test score with one liter per minute more, his expected self-report of limitations would increase by 0.003 if everything else remains constant. The country coefficients decrease further after adding peak flow with the exception of Austria,

9 Sweden, Netherlands, Denmark, and Greece. The lowest and highest coefficients are still in Switzerland and Poland (-2.140 and 1.975). The fifth model adds the chair stand test as a variable to the regression. Chair stand shows a highly significant coefficient of -0.035. So, if a participant were to have a test score with one second more, his expected self-report of limitations would decrease by 0.035 if the

• ther socio-demographic characteristics as well as the grip strength and peak flow performance

remain constant. Chair stand has a minimal higher influence than peak flow, whereas the influence of grip strength the highest of all three performance-based functionality test is. Taking a closer look at this last model, a highly significant but small influence

• f age cannot be denied. If a participant would be one year older, assuming a consistence of all
• ther factors, he is expected to have a .062 higher self-reported limitations score. The gender

influence, however, is bigger. If the participant is a woman, she is expected to have a .769 lower self-reported limitations score than a male respondent. Education has a highly significant influence as well. Therefore, if a participant has one more year of education, he is expected to have a .149 lower self-reported limitations score. Country coefficients experience mainly increases (Austria, Sweden, Netherlands, Spain, Denmark, Switzerland, Belgium and Ireland). The coefficients of Italy, France, Greece, Czech Republic and Poland decrease. Overall, most of the countries show a significant coefficient. However, Austria, Sweden, Belgium, Czech and Ireland do not. The lowest one can be found among Swiss people (-1.498), whereas Poland provides the highest coefficient (1.454). Having a look at the other coefficients, the Southern countries (Spain and Italy) have very low coefficients (-1.410 and -1.031), whereas Denmark, e.g., shows one of -0.501. This means that a Danish respondent is expected to have a 0.501 lower self-reported functionality score than a German one but still higher than a respondent from the South. The final regression model 5 was done with 24.316 observations instead of 30,014

• bservations in the first 4 models. The missing observations are caused by the huge amount of

10 missings in the chair stand test variable. To interfere as less as possible with the number of

• bservations, the chair stand test was added last into the regression models.

The pseudo R2 is highly significant throughout the regressions. In the first model 6.32%

• f the variation of the dependent variable is explained. 8.64% of all variance in the fourth model

are explained when the grip strength and peak flow test results are taken into account next to age, sex, country of origin and education. So, the Pseudo R2 increased. In the last model the value is 0.0809. Therefore, the Pseudo R2 decreased when the chair stand test results are considered. If we change the reference category to Italy (table 3), only Spain and France do not reach statistical significance. With Italy as a reference category, all other countries would have a higher coefficient than Italy, except Switzerland. Austria, Germany, Belgium, Czech and Poland have the highest coefficients of all countries (> 1.00). Conclusion The first question addressed in this paper asked if there statistically significant differences in self-reported functionality among European countries. As shown, there are indeed statistically significant differences in self-reported functionality among the European countries. Our results show that even if age, sex, education and 3 performance-based test results are taken into account, country-specific differences in self-reports of limitations in functioning are still

• present. Those differences depend though on which reference is chosen. A remarkable change

happens though when accounting for different socio-economic drivers: The often in previous research observed North-South increase in reporting limitations is reversed in our analysis. If age, sex, education and 3 performance-based test results are taken into account, Southern European countries report less limitations in functioning than Northern European countries. The second question of our research was if differences in functioning ability as evidence

11 by performance-based measurement explain the differences in self-reports of functioning differences across European countries. As our results showed, the performance-based tests had a highly significant influence on self-reported limitations in functioning. The performance- based measures influenced in different ways the country-specific coefficients. Grip strength lead to an overall decrease of the coefficients throughout the European countries. Peak Flow lead to a further decrease in the majority of the countries. Chair stand though caused very mixed trends and seemed to mainly cause a negative trend in the coefficients. Nevertheless, the impact

• f this influence seems at least for peak flow and chair stand unclear since it is very marginal.

The impact of grip strength seems to be stronger and is actually one of the more powerful drivers in our results, next to sex and education. An important remark should be made regarding the interpretation of the results. Performance-based tests evaluate basic functional limitations, whereas self-reported question items about limitations ask for a general dependency or need for care. Those are two different dimensions of health. Some further limitations of this study have to be mentioned. The SHARE data set provides a unique set of country-specific elderly data. The individual response rates differ between countries from 73.7% to 93.3% [Börsch-Supan et al. 2013]. Even though we used the provided survey weights, this could have influenced the analysis. The different response rates could have led to a selection bias towards unrepresentative samples. In general, it is possible that people who participated in the SHARE survey had fewer limitations or restrictions than those who did not participate. Because of the cross-sectional nature of our study, we could not examine causality in the observed correlations. To determine explicitly causality further research with longitudinal would be necessary. Even though this study contains several limitations, it still has the potential to add to previous research and give important insights in old age health. Our study compared 14 different European countries in order to examine how self-reported functionality differs in those and to estimate the impact of performance-based tests. This was done with data from

12 2006/2007, the most recent wave of the SHARE data set that contains the maximum number of performance-based measures. Acknowledgements 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 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) and from various national funding sources is gratefully acknowledged (see www.share-project.org). References Andersen-Ranberg, K., I. Petersen, H. Frederiksen, J. P. Mackenbach, K. Christensen (2009) Cross-national differences in grip strength among 50+ year-old Europeans: results from the SHARE study. European Journal of Aging, 6, 3. 227 – 236. Angel, R., G. V. Ostir, M. L. Frisco, K. S. Markides (2000) Comparison of a Self-Reported and a Performance-Based Assessment of Moblity in the Hispanic Established Population for Epidemiological Studies of the Elderly. Research on Aging, 22, 6. 715 – 737. Bago d’Uva, T., O. O’Donnell, E. van Doorslaer (2008) Differential health reporting by education level and its impact on the measurement of health inequalities among older

• Euopeans. International Journal of Epidemiology, 37. 1375 – 1383.

Bean, J. F., D. D. Ölveczky, D. K. Kiely, S. I. LaRose, A. M. Jette (2011) Performance-Based Versus Patient-Reported Physical Function: What Are the Underlying Predictors? Physical Therapy, 91, 12. 1804 – 1811.

13 Börsch-Supan, A., A. Brugiavini, H. Jürges, A. Kapteyn, J. Mackenback, J. Siegrist, G. Weber (2008) First results from the Survey of Health, Ageing and Retirement in Europe (2004-2007). Starting the longitudinal dimension. Mannheim: Mannheim Research Institure for the Economics of Aging (MEA). Börsch-Supan, A., M. Brandt, C. Hunkler, T. Kneip, J. Korbmacher, F. Malter, B. Schaan, S. Stuck, S. Zuber (2013) Data Resource Profile: The Survey of Health, Ageing and Retirement in Europe (SHARE). International Journal of Epidemiology DOI: 1093/ije/dyt088. Börsch-Supan, A. (2016) Survey of Health, Ageing and Retirement in Europe (SHARE) Wave

• 2. Release Version: 5.0.0. SHARE-ERIC. Data set. DOI: 10.6103/SHARE.w2.500

Brothers, T. D., O. Theou, K. Rockwood (2014) Do Performance-based Health Measures Reflect Differences in Frailty Among Immigrants Age 50+ in Europe? Canadian Geriatrics Journal, 17, 3. 103 – 107. Coman, L., J. Richardson (2006) Relationship between Self-Report and Performance Measures of Function: A Systematic Review. Canadian Journal on Aging, 25, 3. 253 - 270. Crimmins, E. M., J. K. Kim, A. Solé-Auró (2011) Gender differences in health: results from SHARE, ELSA and HRS. European Journal of Public Health, 21, 1. 81 – 91. Diekmann, A. (2011) Empirische Sozialforschung. Grundlagen. Methoden. Anwendungen. 5.

• Auflage. Hamburg: Rowohlt Taschenbuch Verlag. 242 f..

Gill, T. M. (2010) Assessment of Function and Disability in Longitudinal Studies. Journal of the American Geriatrics Society, 58, 2. 308 – 312. Hair, J. F. Jr., W. C. Black, B. J. Babin, R. E. Anderson (2010) Multivariate Analysis – A Global Perspective. Boston: Pearson. 426. Hairi, F. M., J. P. Mackenback, K. Andersen-Ranberg, M. Avendano (2010) Does socio economic status predict grip strength in older Europeans? Results from the SHARE study in non-institutionalised men and women aged 50+. Journal of Epidemiology and Community Health, 64. 829 – 837.

14 Kapteyn, A., J. P. Smith, A. van Soest (2007) Vignettes and Self-Reports of Work Disability in the United States and the Netherlands. The American Economic Review, 97, 1. 461 - 473. Kempen, G. I. J. M., M. J. G. van Heuvelen, R. H. S. van den Brink, A. C. Kooijman, M. Klein, P. J. Joux, J. Ormel (1996) Factors affecting Contrasting Results between Self- reported and Performance-based Levels of Physical Limitations. Age and Ageing, 25. 458 – 464. Kempen, G. I. J. M., N. Steverink, J. Ormel, D. J. H. Deeg (1996) The Assessment of ADL Among Frail Elderly in an Interview Survey: Self-report versus Performance-Based Tests and Determinants of Discrepancies. Journal of Gerontology: Psychological Sciences, 51B, 5. 254 – 260. Roedersheimer, K. M., G. F. Pereira, C. W. Jones, V. A. Braz, S. A. Mangipudi, T. F. Platts Mills (2016) Self-Reported Versus Performance-Based Assessments of a Simple Mobility Task Among Older Adults in the Emergency Department. Annals of Emergency Medicine. An International Journal, 67, 2. 151 – 156. Sager, M. A., N. C. Dunham, A. Schwantes, L. Mecum, K. Halverson, D. Harlowe (1992) Measurement of Activities of Daily Living in Hospitalized Elderly: A Comparison of Self-Report and Performane-Based Methods. Journal of the American Geriatrics Society, 40, 5. 457 – 462. Seidel, D., C. Brayne, C. Jagger (2011) Limitations in physical functioning among older people as a predictor of subsequent disability in instrumental activities of daily living. Age and Ageing, 40. 463 – 469. Van den Brink, C. L., M. Tijhuis, S. Kalmijn, N. S. Klazinga, A. Nissinen, S. Giampaoli, P. Kivinen, D. Kromhout, G. A. M. van den Bos (2003) Self-reported disability and its association with performance-based limitation in elderly men: a comparison of three European countries. Journal of the American Geriatrics Society, 51, 6. 782 – 788. Verropoulou, G. (2009) Key elements composing self-rated health in older adults: a comparative study of 11 European countries. European Journal of Aging, 6. 213 – 226.

15 Table 1: Weighted sample characteristics by country Country Total N Mean self-reported limitations Mean Age Percentage Female Mean education Peak Flow Grip Strength Chair Stand Austria 571 1.9 65.4 .53 8.7 330 35.6 22.5 Germany 7,400 1.7 64.6 .53 12.5 330 36.3 22.0 Sweden 822 1.5 65.3 .51 11.3 390 36.6 21.2 Netherlands 1,398 1.4 63.6 .53 11.2 348 35.8 18.8 Spain 3,332 1.9 64.5 .54 8.0 296 30.9 24.2 Italy 4,757 2.0 64.8 .54 8.2 263 32.6 36.1 France 4,606 1.7 65.3 .54 11.3 306 33.7 26.3 Denmark 482 1.4 64.2 .52 13.1 362 34.9 14.5 Greece 943 2.0 65.0 .52 8.5 311 33.3 28.8 Switzerland 643 .9 64.8 .53 11.3 348 35.9 15.4 Belgium 934 1.8 65.1 .54 11.8 308 35.2 21.2 Czech Republic 913 1.8 63.6 .55 12.3 302 35.9 23.2 Poland 2,987 3.3 62.8 .56 9.4 268 33.2 32.4 Ireland 236 1.7 63.3 .53 12.1 294 33.4 22.8 Total 30,024 1.9 64.5 .54 10.5 307 34.2 25.9

16 Table 2: Tobit regression models with Germany as reference

Variable Model 1 Model 2 Model 3 Model 4 Model 5 Age .220*** .195*** .121*** .111*** .062*** Sex 1.884*** 1.704***

• .966***
• 1.110***
• .769***

Country Austria .492*

• .246
• .236
• .194

.018 Germany Sweden - .539**

• .775***
• .701***
• .540**
• .288

Netherlands - .549**

• .824***
• .980***
• .941***
• .499**

Spain .122

• .850***
• 1.643***
• 1.651***
• 1.410***

Italy .767**

• .129
• .621**
• .728***
• 1.031***

France - .131

• .372
• .672***
• .699***
• .794***

Denmark - .788***

• .675***
• .990***
• .913***
• .501**

Greece .907*** .082

• .321
• .304
• .402*

Switzerland - 1.999***

• 2.214***
• 2.04***
• 2.140***
• 1.498***

Belgium .169 .049

• .074
• .112

.099 Czech Republic .588** .538** .400* .317 .288 Poland 3.271*** 2.609*** 2.096*** 1.975*** 1.454*** Ireland .434 .333

• .242
• .322
• .305

Education

• .202***
• .176***
• .167***
• .149***

Grip strength

• .161***
• .153***
• .114***

Peak Flow .003*** .002*** Chair stand

• .035***

Constant

• 16.014***
• 11.802***
• .019

1.114 1.401 Pseudo R2 .0632*** .0681*** .0849*** .0864*** .0809*** N 30.024 30.024 30.024 30.024 24.316

17 Table 3: Tobit regression models with Italy as reference

Variable Model 1 Model 2 Model 3 Model 4 Model 5 Age .220*** .195*** .121*** .111*** .062*** Sex 1.884*** 1.704***

• .966***
• 1.110***
• .769***

Country Austria - .275

• .117

.385 .535* 1.050*** Germany - .767*** .129 .621** .728*** 1.031*** Sweden - 1.307***

• .646**
• .080

.189 .743*** Netherlands - 1.316***

• .694**
• .359
• .213

.532** Spain - .645**

• .721***
• 1.022***
• .923***
• .379

Italy France - .898***

• .242
• .051

.030 .237 Denmark - 1.555***

• .546*
• .369
• .185

.530** Greece .140 .211

• .300

.425* .630*** Switzerland - 2.766***

• 2.085***
• 1.583***
• 1.412***
• .467*

Belgium - .598** .178 .547** .616** 1.130*** Czech Republic - .179 .667** 1.021*** 1.045*** 1.319*** Poland 2.504*** 2.738*** 2.717*** 2.704*** 2.485*** Ireland - .333 .462 .379 .406 .727** Education

• .202***
• .176***
• .167***
• .149***

Grip strength

• .161***
• .153***
• .114***

Peak Flow

• .003***
• .002***

Chair Stand .035*** Constant

• 15.247***
• 11.931***
• .640

.386 .370 Pseudo R2 .0632*** .0681*** .0849*** .0864*** .0809*** N 30.024 30.024 30.024 30.024 24.316

18 Graph 1: Adjusted Coefficients Tobit Regression by country