Regional variation in the utilization of nursing home care in Germany

Herr, Annika; Lückemann, Maximilian; Saric-Babin, Amela Article — Published Version Regional variation in the utilization of nursing home care in Germany The European Journal of Health Economics Provided in Cooperation with: Springer Nature Suggested Citation: Herr, Annika; Lückemann, Maximilian; Saric-Babin, Amela (2024) : Regional variation in the utilization of nursing home care in Germany, The European Journal of Health Economics, ISSN 1618-7601, Springer, Berlin, Heidelberg, Vol. 26, Iss. 5, pp. 757-776, https://doi.org/10.1007/s10198-024-01732-9 This Version is available at: https://hdl.handle.net/10419/323293 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) The European Journal of Health Economics (2025) 26:757–776 https://doi.org/10.1007/s10198-024-01732-9 ORIGINAL PAPER Regional variation intheutilization ofnursing home care inGermany AnnikaHerr1,2,3 · MaximilianLückemann1,2 · AmelaSaric‑Babin4 Received: 11 January 2024 / Accepted: 16 October 2024 / Published online: 25 November 2024 © The Author(s) 2024 Abstract Approximately 32 percent of individuals aged over 64years old, with care needs, are residing in nursing homes in Germany. However, this percentage exhibits significant regional disparities, ranging from under 15 percent in certain counties to over 50 percent in others. The purpose of this study is to elucidate the underlying factors explaining this regional variation in nursing home utilization. We employed comprehensive administrative data encompassing the entire elderly care-dependent population and all nursing homes. Our analytical approach involves the use of linear regression models at the county level, accounting for an extensive array of control variables and fixed effects. Additionally, we analyzed regional dependencies by applying spatial lag models. In summary, our model successfully predicts up to 73 percent of the observed regional variation in nursing home utilization. Key factors include care needs, the presence of informal care support and the supply of professional care. Spatial dependencies can be detected but exhibit a minor influence on these variations controlling for care needs. Noteworthy, enabling factors, such as a region’s wealth or rurality, have a very limited impact in a country with a generous social insurance system that covers care for those with limited financial resources. Keywords Regional variation· Long-term care· Spatial panel data models· Nursing home JEL Classification I11· I18· C23 Introduction When it comes to choosing the best type of care, in many countries individuals with care needs have three options: (i) informal care by family members, or (combined with) (ii) formal care by home care providers in their homes, or (iii) nursing home care (NHC). In Germany, home care is prioritized over NHC as per the long-term care (LTC) system’s organizational principle, aiming to keep care recipients in their familiar environment for as long as possible (SGB XI §3). Despite the national regulation of the mandatory LTC insurance, there is significant regional variation in the utilization of nursing home care (NHC) in Germany. The unadjusted mean share varies from below 15 percent to more than 50 percent of all people over 65 years old and in need of care living in nursing homes (NHs) across the more than 400 German counties, with an average of 32 percent. Understanding the disparities in NH utilization is crucial due to substantially higher public and private expenditures for NHC compared to home care or informal care. In 2021, the annual costs for NH were, on average, e21, 000 per person compared to e9, 400 for home care [1]. Furthermore, since demand for LTC is increasing, understanding the regional differences is essential for shaping future LTC policies. Numerous papers have explored regional heterogeneity in healthcare utilization. One prominent example is Finkelstein etal. [2], who explore patient migration and find that a significant proportion of the variation in healthcare utilization can be attributed to demand factors such as preferences and health, besides the major impact of the place-specific supply factors that have been previously * Annika Herr annika.her[email protected]ver.de Maximilian Lückemann maximilian.luec[email protected]ver.de Amela Saric-Babin [email protected] 1 Institute ofHealth Economics (IHE), Leibniz Universität Hannover, CHERH Und CINCH, Königsworther Platz 1, 30167Hannover, Germany 2 CINCH, Universität Duisburg-Essen, Duisburg, Germany 3 Düsseldorf Institute forCompetition Economics (DICE), Hei nrich-Heine-Universität, Düsseldorf, Germany 4 Frankfurt, Germany 758 A.Herr et al. identified. Cutler etal. [3] argue that in the physician market, supply-side characteristics play a more substantial role compared to demand in explaining the regional heterogeneity in expenditures in the US. Godøy and Huitfeldt [4] highlight the influence of socioeconomic factors in explaining regional variation in healthcare utilization and mortality in Norway. Berger and Czypionka [5] show that demand-side factors explain most of the variations in magnetic resonance imaging across medical practices. Reich etal. [6] explore disparities of regional healthcare expenditures in Switzerland and present evidence for its correlation with supply-side densities and socioeconomic factors. Regarding Germany, Augurzky etal. [7] examine the regional differences in the utilization of hospitals, Kopetsch and Schmitz [8] analyze the variation in the usage of ambulatory physician services, and Göppfahrt etal. [9] focus on total healthcare expenditures. Ozegowski and Sundmacher [10] analyze the discrepancy between regional needs and the utilization of outpatient care. They identify supply factors as the primary contributors to this gap. Some of the studies apply spatial autoregressive models and show that correlations are significant but small across counties. In contrast, Felder and Tauchmann [11] generate district-level efficiency scores in health production and highlight the importance of accounting for spatial dependence to analyze the strong effect of federal state-specific regulations on the district’s efficiency. All studies, as is common in this literature, present correlations between the outcomes and the regional explanatory factors rather than causal statements. One prominent exception is Salm and Wu¨bker [12] who utilize individual exogenous patient migration as an instrument to investigate regional differences in ambulatory care utilization. They identify demographics and patient characteristics as being most relevant, in contrast to institutional differences. In terms of LTC provision, Pilny and Stroka [13] examine how the regional availability of NHs affects elderly care decisions using a discrete choice setting where individuals can choose between four different types of formal and informal care. Their study employs resident-level administrative data obtained from a large German health insurance. They find that the decision to choose NHC is significantly driven by the regional supply of NH beds. Mennicken etal. [14] focus on the differences in remuneration rates among NHs in North Rhine-Westphalia. They find that approximately 70 percent of the regional price differences can be explained, the largest part by regional negotiation styles between NHs and the different LTC insurance providers. Lastly, Duell etal. [15] examine the regional variation in the eligibility of publicly financed home care in the Netherlands, which is mainly driven by the place of residence rather than patient experiences. Our study is the first to analyze spatial variations in NHC utilization. We contribute to the existing literature by integrating the healthcare services utilization model proposed by Andersen and Newman [16]. The model comprises three categories: individual determinants (needs, predisposing factors, and enabling factors), health services supply, and societal determinants. We adjust the model for the German LTC setting. We distinguish between individual care needs, the existence of informal care (predisposing), wealth-related factors (enabling), supply measures including prices, and societal determinants. We use comprehensive data combining the German Care Statistic with regional socioeconomic and demographic data at the county level from 2007 to 2019 (biannually). The Care Statistic of the German Statistical Offices of the Länder comprises the entire German care-dependent population and all care facilities. Our methodological approach follows the small area variation studies by Cutler and Sheiner [17] on healthcare expenditures, successfully applied to other German healthcare markets by Augurzky etal. [7], Kopetsch 1. caregiving needs e.g., care levels, age 2. predisposing factors e.g., possibilities of inf. care 3. enabling factors e.g., wealth, prices 2.1 Individual determinants e.g., personnel, equipment, materials, geographic distribution of resources and procedures 2.2 Health services supply e.g., technology available to physicians and behavioral norms 2.3 Societal determinants Fig. 1 Determinants of the Andersen-Newman model of healthcare utilization Illustration of the three indicator groups derived from the Andersen-Newman model of healthcare utilization: (1) individual determinants, (2) health services supply, and (3) societal determinants 759Regional variation intheutilization ofnursing home care inGermany and Schmitz [8], and Ozegowski and Sundmacher [10]. We estimate ordinary least squares (OLS) models with regionalfixed effects and wave-fixed effects to uncover the variation of our outcome variable. To account for spatial dependencies in utilization and regional shock spillovers inspired by Gupta etal. [18] and Ozegowski and Sundmacher [10], we additionally apply spatial autoregressive (lag) models. Overall, our model achieves a good measure of fit explaining 73 percent of the variation in NH utilization across the more than 400 German counties, of which almost 64 percent can be attributed to care needs (explaining 47 percent of the variation). This is more than previous studies have found for other markets. Regional predisposing indicators, such as informal care opportunities measured by the female workforce and demographics, account for an additional 8–12 percentage points (depending on the order of inclusion). Regional enabling factors, including wealth and rurality, make a very small contribution, explaining an additional 1–2 percentage points of the variation. The supply of healthcare adds 5–10 percentage points when controlling for care needs, which is much less than in previous studies. Regional and wave-fixed effects finally explain 7 percentage points of the variation (added at last to the model). Furthermore, we identify a small yet significant presence of spatial dependencies that do not alter the main conclusions. The findings have important policy implications since they demonstrate that besides the care needs, the degree of informal care support correlates highly with the demand for NHC. Thus, policymakers could either stimulate informal caregiving in some areas by reducing the double burden of work and caring or, in turn, increase the availability of NHC to reduce the need for informal care provision. It is also important to note that wealth and income cannot explain the variation even when not controlling for the supply of formal care that may be correlated. Section "A model of health services utilization" discusses the model of health services utilization we apply to this setup, followed by Section "Data and descriptive statistics”, which presents the data and descriptive statistics. We introduce the estimation strategy in Section "Estimation strategy” and then present our results in Section "Results”. In Section "Discussion and conclusion”, we discuss our findings and conclude. A model ofhealth services utilization We examine the regional disparities in the utilization of NHC by employing a modified version of the AndersenNewman model of healthcare utilization [16]. The model distinguishes between Individual determinants of utilization individual determinants, Health services supply supply of LTC services, and Societal determinants societal determinants of utilization. All parameters are aggregated at the county level. The individual characteristics from the care statistics are based on all elderly individuals (aged 65 or older) receiving any contribution from the LTC insurance system (across the three care types if not indicated differently). For an overview, all variables are grouped and defined briefly, including sources, in Table5in the appendix. Individual determinants ofutilization As individual determinants of utilizing healthcare services are multi-faceted, we specify three groups of personal characteristics: care needs, predisposing factors, and enabling factors. Care needs correspond to individual disabilities or health status. With regard to care needs, we include the average county-level care level it corresponds to. German regulations distinguish between three levels of LTC severity in our data.1 The need for caregiving arises if individuals require assistance with the daily living activities due to advanced age, healthor mental-related problems for a minimum of six months (SGB XI §14), which is assessed by external bodies of the mandatory LTC insurance, the so-called Medical Review Boards (Medizinischer Dienst der Krankenkassen [MDK] in German). They also set rules to negotiate prices between the providers and the statutory LTC insurances and communities and monitor quality on behalf of the health insurance providers (Fig.1). They operate at the federal state level except for North Rhine-Westphalia (separated into North Rhine and Westphalia-Lippe), Hamburg (combined with Schleswig-Holstein), and Berlin (combined with Brandenburg).2 People who have been assigned a care level receive financial support from their LTC insurance (see respective allowances in Table6in the appendix). Level 1 indicates moderate needs while higher levels correspond to more severe health problems. Care recipients in care level 3 are disproportionately highly represented in NHs [19]. Age correlates with the emergence of frailties and therefore serves as a proxy for healthcare needs and patterns of medical care use [16]. We split the elderly care recipients (across all care types) into three groups: 65–74, 75–84 and 85 and over, where the lowest age group is the reference category. This follows the idea that older care recipients are more likely to use NHC than their younger peers [19, 20]. Predisposing characteristics capture the surrounding socio-demographic conditions. We employ several 1 In 2017, two levels were added, which we map to the former system the following way: the new care grades 1 and 2 correspond to level 1, care grade 3 to level 2, and care grades 4 and 5 to level 3. 2 The MRB is an independent non-profit organization providing socio-medical specialist advice to the German Statutory Health Care and LTC Insurance. An overview of federal state LTC regulations is available at http:// www. biva. de/ geset ze/ laend erheimg esetze/, last accessed in July 2024. 760 A.Herr et al. Table 1 Descriptive statistics We report descriptive statistics for all German elderly care-dependent individuals (aged 65+) entitled to public and private long-term care insurance allowances aggregated at the county level for the years 2007, 2009, 2011, 2013, 2015, 2017 Sources: Research Data Centre (RDC) of the Federal Statistical Office and Statistical Offices of the Länder, Care Statistic, 2007 and 2017 (DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0–https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0) [32]; INKAR database of the Federal Office for Regional Planning [26]; IAB administrative vacancy information from the Research Institute of the Federal Employment Agency [35]; Transparency report cards from the BKK comparison engine for stationary LTC https://pflegefinder. bkk-dachverband.de/. Own calculations *Labor force participation is a share of males/females in the age group 15–65 in the active workforce **Rurality is defined as a share of county’s population living in municipalities with less than 150 residents per km 2 ***OOP refers to a price negotiated for each nursing home net of the allowance paid by the mandatory LTC insurance ****Vacancy ratio is the ratio of open positions to the total number of budgeted positions ***** Nursing care quality is an indicator based on the quality report cards of the Medical Review Boards (MRB), between zero (very poor quality) and one (excellent quality) (see A1 for details in the Appendix). 2007-2017 Mean sd p1 p99 Nursing home utilization  Elderly care recipients in nursing homes [%] 31.89 7.10 17.93 49.97 Caregiving needs  Elderly care recipients across all care types  Share in care level 2 35.06 5.60 27.17 50.81  Share in care level 3 11.15 3.04 5.06 19.40  Share aged 75-84 43.69 3.89 33.74 52.55  Share aged 85+ 43.69 3.89 33.74 52.55 Predisposing  Labor force participation* [%]  Female 76.69 4.90 67.00 85.25  Male 84.59 4.26 72.30 92.60  Inhabitants by age group [%]  Age group 35–49 27.17 2.19 22.60 32.40  Age group 50–64 21.34 2.54 16.75 27.60  Age group 65–74 11.08 1.42 8.20 14.80  Age group 75–84 7.68 1.32 5.30 11.50  Age group ≥ 85 2.56 0.46 1.60 3.70  Avg. life expectancy at age 60[years] 83.41 0.678 81.97 84.98 Enabling  GDP per capita [1000 EUR] 31.92 14.24 15.67 89.87  Use of social assistance for elderly [%] 19.28 12.67 3.80 61.80  Avg. pension [EUR] 833.83 84.02 659.50 1036.75  Monthly available household income [EUR] 1681.77 245.76 1228 2419  Share rurality** 0.29 0.29 0.00 1.00  Communal debts per capita [EUR] 1639.02 1365.52 0.00 6582.45 LTC supply  Hospital beds [per 10,000 inh.] 6.44 3.84 0 19.29  Home care facilities [#] 34.49 41.26 7.00 180.00  Beds per residents (occupancy rate) in NH 0.89 0.06 0.74 0.99  Single-room share in NH [%] 61.12 11.06 36.02 85.18  Nurse vacancy ratio**** 0.035 0.02 0.01 0.11  Personnel per resident, share 0.62 0.09 0.39 0.82  OOP*** care level 1 [EUR] 1436.82 316.40 719.29 2289.61  OOP*** care level 2 [EUR] 1572.76 334.09 809.01 2295.09  OOP*** care level 3 [EUR] 1788.19 368.65 1015.41 2535.70 Observations 2,221 761Regional variation intheutilization ofnursing home care inGermany variables as proxies for the existence of informal care support: the shares of men and women in the active workforce, the regional demographics captured by the share of inhabitants in each age cohort, each with a different probability of serving as informal caregivers, and the average life expectancy at age 60. Labor force participation is defined as the share of men (or women) aged between 15 and 65 years in the active workforce. The effect of employment on the choice of the type of care is ambiguous. Employment reduces the capacity to provide informal care; yet flexible work arrangements (part-time, mini-jobs) can help reconcile work and care duties [21]. The propensity of daughters to provide informal care is generally higher than that of sons [22–25]. We expect that a higher share of women in the active workforce is associated with higher use of NHC, while the effect of men’s participation is not clear a priori. For example, women could partly substitute work with caregiving, while their spouses compensate for the earnings foregone by working more. Around 48 percent of women in the active workforce in 2019 were part-time employed, while the corresponding share of men was only 11.2 percent [26]. We include the share of the population aged below 34 as reference category and add 35 to 49, 50 to 64, 65 to 74, 75 to 84, and 85 and older. The majority of caregiving relatives (children and their spouses) fall into the age span of 50–64. Augurzky etal. [7] suggest that most informal caregivers are between 55 and 69 years of age, which we approximate with the available data (50–65 years old). We also include life expectancy at age 60 and postulate that higher life expectancy is associated with better health in older ages, resulting in lower demand for NHC. However, having access to NHC might also increase life expectancy for the oldest care recipients. Enabling characteristics predominantly cover the propensity of the affordability and accessibility of NHC. Here, we include GDP per capita, household income, average pension for seniors, rurality, the utilization of social aids for the elderly, and communal debts. Rurality serves as a proxy for travel times as distance influences facility choices [7, 27–29]. Rurality is measured as the share of the county’s population living in municipalities with fewer than 150 inhabitants per square kilometer. We postulate that NHC in Germany is utilized more frequently in urban than in rural areas. NHC is considered to be a normal good, meaning that demand is negatively affected by higher prices and positively by income and wealth. Therefore, we include various measures of income and impoverishment, such as the amount of social aid for the elderly and communal debts. Living in an NH is relatively expensive, with an average co-payment of e1, 620 per month (Table1). In line with this, Bakx etal. [30] report that being in the bottom income quartile in Germany decreases the probability of using any formal LTC. Similarly, Augurzky etal. [4] and Eibich and Ziebarth [31] find that a higher income positively impacts the utilization of hospital and physician services. Health services supply Health services supply encompasses the resources and organization of healthcare delivery. This category includes personnel, equipment, materials, resources, and procedures employed once a person in need of care becomes part of the LTC system. To approximate the regional supply of healthcare services, we consider various provider characteristics. Firstly, we include the number of hospital beds per 10,000 inhabitants, which can potentially serve as a temporary alternative for NHC, especially following surgical interventions in seniors [32]. Secondly, we postulate that the availability of home care may also contribute to delaying the utilization of NHC. To assess the supply of NHC, we include the average single-room share, the personnel vacancy ratio, the occupation ratio, and the personnel-to-resident ratio. We include NH prices by care level 1 to 3 as the log of the out-of-pocket payment, i.e., the share of the price that is not covered by the LTC insurance (either financed privately out-of-pocket or, if not feasible, covered by the social insurance system). Lastly, we control for the NH’s quality by including a quality indicator (see the appendix for the definition). Societal determinants Finally, societal determinants of NH choice include the technology available to physicians and behavioral norms. Utilization relative to German average 1050-5-10 Brandenburg Hesse Mecklenburg Western Pomerania Berlin Bremen Thuringia Saarland Rhineland-Palatinate North Rhine-Westphalia Germany Saxony-Anhalt Lower Saxony Saxony Baden-Wuerttemberg Hamburg Bavaria Schleswig-Holstein Percentage difference Fig. 2 Nursing home (NH) utilization across federal states compared to the German average. Data source: Statistical Offices of the Länder, Care Statistic, 2007 to 2017 (DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0–https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0). Note: This figure shows average federal-state deviation from our main dependent variable mean over time from 2007 to 2017. The average share of care recipients moving into NHs is appr. 31%. Deviations are provided in percentage points 762 A.Herr et al. Societal determinants of utilization are generally not observable [31]. We hypothesize that the regional fixed effects capture not only the geographic differences in technology but also differences in behavioral norms. Data anddescriptive statistics We exploit three data sources. First, we use the German care statistics from the Federal Statistical Offices and the Statistical Offices of the Länder at the Research Data Center (RDC) of Hannover (“Care Statistics”). The bi-annual statistic spans over six waves from 2007 to 2017, reflecting the German care situation on December 15 every two years. The data comprise all care recipients entitled to public or private LTC insurance allowances in Germany. This includes both informal and professional care recipients, where the latter are differentiated between NHC and home care. The data also cover all NHC and home care providers, e.g., the number of available places, the personnel, the type of caregiving services offered, and the fees to be paid to the care facility for caregiving, accommodation, and meals. To focus on regional variation rather than individual determinants of NH entry, we aggregate the data at the county level. Second, we supplement the care statistic with indicators of regional and urban development, such as socio-economics, demographics, and health care supply, at the county level provided by the Federal Office for Building and Regional Planning (INKAR) [7] (in German: Indikatoren und Karten zur Raumund Stadtentwicklung [INKAR], https:// www. inkar. de/.) Third, we add labor market information from the German Institute for Employment Research, more specifically, the number of vacancies for nurses in geriatric care per county and year [41]. In our final sample, we excluded all people under 65 years of age in need of care and also all care facilities that provide non-elderly care (e.g., care for kids or psychiatric care) or facilities that provide only short-term care or only day or night care. In compliance with the data security rules enforced by the research data center, we aggregated neighboring counties with fewer than three NHs per ownership type and county, which gives us 374 (2007) and 367 (2017) regions covering all 400 counties across the 16 federal states (2221 observations over six waves). The dependent variable is defined as the county’s share of elderly care recipients aged 65 or older living in NHs as opposed to living at home (receiving only informal care or home care support). Figure2 demonstrates regional disparities in the utilization of NHC across federal states. NHC is most frequently used in Schleswig-Holstein, Bavaria, and Hamburg. In Schleswig-Holstein, the proportion of care recipients in NHs is consistently 9–10 percentage points higher than the national average. Conversely, Brandenburg, Hesse, and Mecklenburg Western Pomerania fall 5–6 percentage points below the average. The appendix provides maps that visualize the regional variation in LTC utilization across care types and its changes over time. From these maps, we can discern certain patterns. Figure4 illustrates the variation in NHC across counties in 2007 and 2017. NHC appears to be more prevalent in the northern and southern regions of Germany. For the two other care types, there is a higher utilization of home care in the northeast (Figure5) and greater reliance on informal care in the southwest (Figure6). Table1 summarizes all variables included in our model. For detailed variable definitions compare Table5 in the appendix. The table shows that there is strong regional variation across all variables when looking at the lowest and highest percentile. Estimation strategy Multivariate linear regressions toexplain variation inNHC Our estimation approach follows Cutler and Sheiner [17], Kopetsch and Schmitz [8], and Berger and Czypionka [5] and allows us to net out the variation that is due to systematic differences between the counties. We gradually add groups of explanatory variables to the regression and infer their explanatory power from the changes in goodness-of-fit measures. Our preferred order of inclusion corresponds to the models presented in Section "A model of health services utilization". The simple linear regression equation is given as where yct is the share of care recipients in NHC in county c and wave t. X varies across models depending on the control variables included in groups. We start with care needs and subsequently include predisposing, enabling, and supply factors. Finally, in the full model, we add wave-fixed effects τt and regional fixed effects R to account for unobservable factors at the relevant administration level r, which are the Medical Review Boards (MRB or MDK in German) since they set the rules to negotiate prices and monitor quality on behalf of the health insurance providers as discussed above. µct indicates the IID disturbances term. The standard errors are clustered by region and time (MRB times year). Since the explanatory power of each variable block varies based on the sequence of inclusion, we also specify (1) yct = Xct � 𝛽 + Rr + 𝜏t + 𝜇ct 763Regional variation intheutilization ofnursing home care inGermany alternative sequences for robustness analyses (compare section “Alternative Ordering”). Spatial auto‑correlation acrosscounties As German counties cluster in bigger regions, e.g., federal states, there may be correlations in healthcare use across counties. That is why, in a second step, we account for confounding spatial dependencies by employing spatial regression models and test whether the coefficients of the explanatory variables change compared to the non-spatial linear regressions. In spatial interaction-based models, collective behaviors and aggregate patterns are assumed to emerge from the interaction of agents across social, economic, and geographic dimensions [35, 36]. Interaction can be (a) endogenous, where the group causally influences individual behavior, (b) exogenous, where individual behavior varies with exogenous characteristics of the group, or (c) correlated, where similar behavior is due to similar individual characteristics and institutional environments [37]. The underlying idea is that actions chosen by one individual influence the constraints, expectations, and preferences in her reference group [38]. Endogenous interactions in NHC may arise from cultural factors. For example, the high use of NHC may increase its broader societal acceptance. In our context, exogenous interaction can be attributed to a wider effect of local developments. For example, the closure of a large NH in one county is likely to boost the demand for NHC in neighboring counties. Negative economic shocks will increase unemployment, which could increase the degree of informal care support. Correlated interactions result from factors that cannot be observed in the data. Examples include a high prevalence of conditions that often precede moving to an NH, such as mental diseases and strokes, or a good quality of care in a particular region. We follow Kopetsch and Schmitz [8] and Gupta etal. [18], who analyze spatial dependencies in other healthcare markets, and assume that the use of NHC follows a spatial autoregressive process. Equations (2) and (3) formally describe the estimation procedure of the spatial autoregressive combined model (SAC), including a spatially lagged outcome and error term: where W originates from the spatial contiguity weight matrix, and 𝜇ct and 𝜖ct are vectors of spatially correlated residuals and IID disturbances, whereby 𝜖ct ∼ N (0, σ2I). We denote Wyjt and Wujt as spatial lags of the dependent (2) yct = 𝜆Wyjt + Xct � 𝛽 + Rr + 𝜏t + 𝜇ct (3) uct = 𝜌Wujt + 𝜖ct, (Moran's I=0.2301 and P-value=0.0010) 3 2 1 0 -1 -2 -2 024 NH share WNHshare Fitted values Spatially laggedNH_share Fig. 3 Moran scatter plot. Data source: RDC of the Federal Statistical Office and Statistical Offices of the Länder, Care Statistic, 2007 and 2017 (DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0– https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0), own calculations. Notes: The Moran scatter plot displays how the selected attribute’s values at each location relate to the average value of the same attribute at neighboring locations. The upper-right (lower-left) quadrant represents cases where both the attribute value and the local average value exceed (lie below) the overall average value, indicating positive spatial autocorrelations. The other two quadrants indicate negative spatial autocorrelations. The dominant groups within these quadrants determine the overall tendency toward positive, negative, or no spatial autocorrelations (see, e.g., lecture notes from Penn State, Project 4: Calculating Global Moran’s I and the Moran Scatter plot, https:// www.eeduca tion. psu. edu/ geog5 86/ node/ 672, last accessed in July 2024) 764 A.Herr et al. Table 2 Regression results on variation in NHC, 2007–2017 Model 1 Model 2 Model 3 Model 4 Model 5 Elderly care recipients in level 2 [%] – 0.0092 – 0.0307 – 0.0504 – 0.0995 0.0003 (0.0590) (0.0477) (0.0511) (0.0658) (0.0687) Elderly care recipients in level 3 [%] 0.5560*** 0.4139*** 0.3959*** 0.1967*** 0.2422*** (0.0969) (0.0513) (0.0802) (0.0729) (0.0555) Elderly care recipients aged 75 – 84 [%] – 1.627*** – 0.7844*** – 0.8873*** – 0.7100*** – 0.4878*** (0.1480) (0.1816) (0.1815) (0.1447) (0.1013) Elderly care recipients aged 85+ [%] – 0.211* 0.8343*** 0.6711*** 0.7135*** 0.7842*** (0.1201) (0.1279) (0.1297) (0.1101) (0.0926) Predisposing Female labor force participation, share 0.2433*** 0.2585*** 0.1116* 0.2746*** (0.0580) (0.0593) (0.0582) (0.0591) Male labor force participation, share – 0.1135*** – 0.1671*** – 0.0949*** – 0.2011*** (0.0414) (0.0384) (0.0336) (0.0313) Inhabitants 35–49, share 0.2404 0.1567 0.4522*** 0.2393 (0.1736) (0.1770) (0.1401) (0.1554) Inhabitants aged 50–64, share – 0.9781*** – 0.9556*** – 0.8012*** – 0.3347*** (0.1210) (0.1260) (0.1376) (0.1032) Inhabitants 65–74, share 0.5286*** 0.3698** 0.2794 – 0.6543*** (0.1835) (0.1724) (0.1994) (0.1846) Inhabitants 75–84, share 1.919*** 1.626*** 1.395*** 1.544*** (0.3374) (0.4323) (0.3940) (0.2699) Inhabitants ≥85 , share – 4.083*** – 3.357*** – 3.256*** – 2.204*** (0.8868) (1.0100) (0.8692) (0.6460) Avg. life expectancy at 60, years – 0.0386*** – 0.0430*** – 0.0369*** – .0325*** (0.0041) (0.0041) (0.0035) (.003) Enabling Log(GDP per capita [1,000 EUR]) 0.0143** – 0.0045 0.0011 (0.0061) (0.0057) (0.0052) Social aids for elderly [%] – 0.0740*** – 0.0185 – 0.0059 (0.0225) (0.0239) (0.0157) Log (Avg. pension) 0.0166 0.0465 0.1589*** (0.0278) (0.0282) (0.0268) Share rurality** – 0.0165* – 0.0237*** – 0.0215*** (0.0083) (0.0072) (0.0061) Log (Available household income) 0.0229 0.0429** 0.0189 (0.0177) (0.0200) (0.0174) Log (Communal debts p.c. [EUR]) 0.0007 0.0010 0.0002 (0.0007) (0.0010) (0.0006) Supply Hospital beds [per 10,000 inh.] 0.0022*** 0.0022*** (0.0004) (0.0003) Single-room share in NH [%] – 0.0388*** – 0.0328** (0.0138) (0.0136) Home care facilities – 0.0002*** – 0.0002*** (0.00005) (0.00005) Occupancy rate in NH – 0.1002*** – .1556*** (0.0317) (0.0280) Nurse vacancy ratio**** – 0.4489*** – 0.3009*** (0.0663) (0.0590) Personnel per resident ratio – 0.0848*** – 0.1263*** 771Regional variation intheutilization ofnursing home care inGermany unadjusted + caregiving needs + predisposing + enabling + supply + fixed effects + spatial corr. dep. var. + spatial lag and error model 0 .5 1 1.5 Fig. 7 Observed-to-predicted ratio of NHC recipients by quartile of observed utilization. Note: We classify the counties into quartiles based on the unadjusted shares of care recipients in NH. blue: lowest 25%, red: 25–49%, green: 50–75%, yellow: highest quartile. The specifications used are as follows: 1. Unadjusted; 2. Adjusted for need; 3. Additionally adjusted for predisposing factors; 4. + enabling factors; 5. + NH supply; 6. + MRB and wave fixed effects; 7 + 8 + spatial correlations. Data source: RDC of the Federal Statistical Office and Statistical Offices of the Länder, Care Statistic, 2007 and 2017 (DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0–https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0). Own calculations Fig. 8 Shares of care recipients in NHC and spatial dependence. Average shares of care recipients in NH for 2007 and 2017 across counties above or below the national average and the sign of its spatial dependence. ( – , +) utilization below mean, positive spatial dependence; ( – , – ) utilization below mean, negative spatial dependence; (+ , – ) utilization above mean, negative spatial dependence; (+ , +) utilization above mean, positive spatial dependence. Data source: RDC of the Federal Statistical Office and Statistical Offices of the Länder, Care Statistic, 2007 and 2017 (DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0–https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0). Own calculations 772 A.Herr et al. Tables See Tables5, 6, 7 Table 5 Description of explanatory variables and sources, aggregated at county level Data sources: Care statistic Care statistic, (DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0–https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0); INKAR database of the Federal Office for Building and Regional Planning (BBR); IAB German Institute for Employment Research; Transparency reports Transparency report cards from the BKK comparison engine for stationary LTC: https://pflegefinder.bkk-dachverband.de/ [last accessed Sept. 2024] Description Datasource NHCutilization Elderly care recipients in NH Share of people in need of care above 65 using NHC Care statistic Care needs Elderly care recipients in care level 1–3 across all care types Share of people in need of care in care levels 2 and 3, compared to care level 1 Care statistic Elderly care recipients aged 75–84 (85+), share Share by age group, reference category: 65–74 aged care recipients Care statistic Predisposing Labor force participation [%] Share of employed individuals in the county per 100 inhabitants of working age (15–65) INKAR Inhabitants, by age groups, [%] Share of total population by age, <34 is the reference category INKAR Avg. life expectancy at 60 [years] Average life expectancy for a 60-year-old person INKAR Enabling GDP per capita [1,000 EUR] Gross Domestic Product (GDP) per inhabitant INKAR  Use of social aid for the elderly [%] Share of the population receiving basic elderly support INKAR Avg. pension [EUR] Average monthly pension payment in euros INKAR Monthly household income per capita [EUR] Average disposable household income in euros per inhabitant INKAR Rurality ** [%] Share of a county’s inhabitants living in municipalities with less than <150 inhab. per km 2 INKAR Communal debts per capita [EUR] Communal debts per inhabitant INKAR LTC supply Hospital beds [per 10,000 inh.] Hospital beds per 10,000 inhabitants INKAR Single room share in NH [%] Average share of single rooms in NH Care statistic Occupancy rate Share of residents per NH bed Care statistic Personnel to resident ratio Share of full-time equivalent personnel to all residents Care statistic Home care facilities [#] Number of home care providers (ambulatory care) Care statistic Nurse vacancy ratio Share of job openings for geriatric care nurses to all budgeted positions IAB Average OOP by care level [EUR]*** Out-of-pocket contribution by care level (=NH price−LTC insurance payment as of Table6) Care statistic Nursing care quality Index of nursing care quality (see Equation4 for further information) Transparency reports 773Regional variation intheutilization ofnursing home care inGermany Table 6 Long-term care insurance funds’ allowance We present the maximum monthly allowance (euros) paid by the German public long–term care insurance for informal care, home health care, and nursing home care. Depending on their contributions, care recipients with private insurance are entitled to higher allowances. Sources: Bundesgesundheitsministerium; Pflegestärkungsgesetze I–III, Pflege-Neuausrichtungs-Gesetz. 2007-2011 (pp. 44-45): https:// www. bunde sgesu ndhei tsmin ister ium. de/ filea dmin/ Datei en/5_ Publi katio nen/ Pflege/ Beric hte/5. Pfleg eberi cht. pdf, 2011– 2015 https:// www. bunde sgesu ndhei tsmin ister ium. de/ filea dmin/ Datei en/5_ Publi katio nen/ Pflege/ Beric hte/6. Pfleg eberi cht. pdf (pp. 116–117) and https:// www. bunde sgesu ndhei tsmin ister ium. de/ minis terium/ meldu ngen/ 2016/ dezem ber2016/ neure gelun gen2017. html for 2017 [last accessed Nov. 2024] 2007 2009 2011 2013 2015 2017* Informal care  Care level I 205 215 225 235 244 316  Care level II 410 420 430 440 458 545  Care level III 665 675 685 700 728 728  Hardship case – – – – – 901 Home health care  Care level I 384 420 440 450 468 689  Care level II 921 980 1040 1100 1144 1298  Care level III 1432 1470 1510 1550 1612 1612  Hardship case 1918 1918 1918 1918 1995 1995 Nursing home care  Care level I 1023 1023 1023 1023 1064 770  Care level II 1279 1279 1279 1279 1330 1262  Care level III 1432 1470 1510 1550 1612 1775  Hardship case 1688 1750 1825 1918 1995 2005 774 A.Herr et al. Table 7 Spatial regression results based on OLS, model 5, 2007–2017, full results ∗p<0.10 , ∗∗ p<0.05 , ∗∗∗ p<0.01 . Standard errors suppressed for ease of presentation Specifications: (1) Spatial Lag Model (SLM), Equation (2), (2) Spatial Error Model (SEM) Equation (3), (3) Spatial Autoregressive Combined Model (SAC), Equations (2) and (3). Sources: Research Data Centre (RDC) of the Federal Statistical Office and Statistical Offices of the Länder, Care Statistics, survey years 2007–2017 (DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0–https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0); INKAR database of the Federal Office for Building and Regional Planning (BBR). Own calculations SLM SEM SAC Caregiving needs Care recipients in care level 2, share 0.0153 0.0114 0.0154 Care recipients in care level 3, share 0.2468*** 0.3443*** 0.3476*** Share LTC recipients aged 75–84 – 0.4886*** – 0.4742*** – 0.4712*** Share LTC recipients aged 85+ 0.7536*** 0.7402*** 0.7225*** Predisposing Female labor force participation, share 0.2801*** 0.2707*** 0.2773*** Male labor force participation, share – 0.2143*** – 0.1975*** – 0.2091*** Share inhabitants 35 −49 0.1545 0.1678 0.1040 Share inhabitants aged 50 −64 – 0.3499*** – 0.3886*** – 0.4043*** Share inhabitants 65 −74 – 0.6831*** – 0.3948** – 0.4215** Share inhabitants 75 −84 1.4752*** 1.2300*** 1.2138*** Share inhabitants ≥85 – 1.9391*** – 1.9298*** – 1.7317*** Avg. life expectancy at 60 – 0.0316*** – 0.0317*** – 0.0315*** Enabling Log (GDP per capita [1,000 EUR]) 0.0031 0.0022 0.0030 Use of social aids for elderly – 0.0017 0.0164 0.0218 Log (Avg. pension) 0.1355*** 0.1612*** 0.1507*** Share rurality** – 0.0234*** – 0.0182*** – 0.0192*** Log (Available household income) 0.0220* 0.0205 0.0220* Log (Communal debts [EUR]) – 0.0001 0.0001 – 0.0001 Supply Hospital beds [per 10,000 inh.] 0.0023*** 0.0019*** 0.0021*** Single room share in NH – 0.0300*** – 0.0339*** – 0.0320*** Ratio ambulatory facilities per care recipient – 0.0002*** – 0.0002*** – 0.0002*** Occupancy rate in NH – 0.1576*** – 0.1507*** – 0.1510*** Vacancy ratio**** – 0.3091*** – 0.3332*** – 0.3334*** Nurses per resident – 0.1187*** – 0.1249*** – 0.1195*** Log(Average OOP*** [EUR], Care level 1) 0.0967** 0.1277*** 0.1237*** Log(Average OOP*** [EUR], Care level 2) – 0.1306** – 0.1056 – 0.1053* (0.056) (0.064) (0.064) Log(Average OOP*** [EUR], Care level 3) – 0.0030 – 0.0553 – 0.0523 Nursing care quality***** 0.0197** 0.0168** 0.0177** Constant 2.3142*** 2.1161*** 2.1749*** Lambda ( 𝜆 ) 0.0060*** 0.0065*** (0.001) (0.002) Rho ( 𝜌 ) 0.0823*** 0.0770*** (0.006) (0.006) MRB FE, Time FE Yes Yes Yes Observations 2,221 2,221 2,221 775Regional variation intheutilization ofnursing home care inGermany Acknowledgements We thank Martin Karlsson and Andreas Schmid as well as participants of the workshop DIBOGS 2017 in Munich, the dggö conference 2022, and the CINCH seminar at Essen for valuable comments. We thank Urban Janisch and Jonas Löbel from the Research Data Centre of the Statistical Offices of Saxony-Anhalt for providing us with the care statistics. We would also like to express our gratitude to Florian Köhler from the Research Data Centre at the Statistical Office of Lower Saxony for his support during our on-site work and data processing. Financial support from the German Ministry of Education and Research (BMBF; grant numbers 01EH1102A and 01EH1602B) and the German Research Foundation (DFG; grant number HE 6825/3-1) is gratefully acknowledged. Funding Open Access funding enabled and organized by Projekt DEAL. Deutsche Forschungsgemeinschaft, HE 6825/3-1, Annika Herr, Bundesministerium für Bildung und Forschung, 01EH1602B, Annika Herr, 01EH1102A, Annika Herr. Data availability The Care Statistics are publicly available and can be accessed via the Statistical Office of the Länder under the contracted restrictions (e.g., on-site use and fees). DOI: https:// doi. org/ 10. 21242/ 22411. 2007. 00. 02.1. 1.0–https:// doi. org/ 10. 21242/ 22411. 2017. 00. 02.1. 1.0. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. References 1. GKV-Spitzenverband: Kennzahlen der sozialen Pflegeversicherung 2023, www. gkvspitz enver band. de/ servi ce/ zahlen_ und_ grafi ken/ spv_ kennz ahlen/ spv_ kennz ahlen. jsp, (2022) 2. Finkelstein, A., Gentzkow, M., Williams, H.: Sources of geographic variation in health care: evidence from PatientMigration. Q. J. Econ. 131, 1681–1726 (2016). https:// doi. org/ 10. 1093/ qje/ qjw023 3. Cutler, D., Skinner, J.S., Stern, A.D., Wennberg, D.: Physician beliefs and patient preferences: a new look at regional variation in health care spending. Am. Econ. J. Econ. Pol. 11, 192–221 (2019). https:// doi. org/ 10. 1257/ pol. 20150 421 4. Godøy, A., Huitfeldt, I.: Regional variation in health care utilization and mortality. J. Health Econ. 71, 102254 (2020). https:// doi. org/ 10. 1016/j. jheal eco. 2019. 102254 5. Berger, M., Czypionka, T.: Regional medical practice variation in high-cost healthcare services. Eur. J. Health Econ. 22, 917–929 (2021). https:// doi. org/ 10. 1007/ s1019802101298-w 6. Reich, O., Weins, C., Schusterschitz, C., Thöni, M.: Exploring the disparities of regional health care expenditures in Switzerland: some empirical evidence. Eur. J. Health Econ. 13, 193–202 (2012). https:// doi. org/ 10. 1007/ s101980110299-x 7. Augurzky, B., Kopetsch, T., Schmitz, H.: What accounts for the regional differences in the utilisation of hospitals in Germany? Eur. J. Health Econ. 14, 615–627 (2012). https:// doi. org/ 10. 1007/ s101980120407-6 8. Kopetsch, T., Schmitz, H.: Regional variation in the utilisation of ambulatory services in Germany. Health Econ. 23, 1481–1492 (2014). https:// doi. org/ 10. 1002/ hec. 3001 9. Göpffarth, D., Kopetsch, T., Schmitz, H.: Determinants of regional variation in health expenditures in Germany. Health Econ. 25, 801–815 (2016). https:// doi. org/ 10. 1002/ hec. 3183 10. Ozegowski, S., Sundmacher, L.: Understanding the gap between need and utilization in outpatient care—the effect of supply-side determinants on regional inequities. Health Policy 114, 54–63 (2014). https:// doi. org/ 10. 1016/j. healt hpol. 2013. 08. 005 11. Felder, S., Tauchmann, H.: Federal state differentials in the efficiency of health production in Germany: an artifact of spatial dependence? Eur. J. Health Econ. 14, 21–39 (2013). https:// doi. org/ 10. 1007/ s101980110345-8 12. Salm, M., Wübker, A.: Sources of regional variation in healthcare utilization in Germany. J. Health Econ. 69, 102271 (2020). https:// doi. org/ 10. 1016/j. jheal eco. 2019. 102271 13. Pilny, A., Stroka, M.A.: Determinants of received long-term care – Individual responses to regional nursing home provisions. Health Care Manag. Sci. Manag. Sci. 19, 326–337 (2016). https:// doi. org/ 10. 1007/ s107290159333-3 14. Mennicken, R., Augurzky, B., Rothgang, H., Wasem, J.: Explaining differences in remuneration rates of nursing homes in Germany. Eur. J. Health Econ. 15, 401–410 (2014). https:// doi. org/ 10. 1007/ s101980130483-2 15. Duell, D., van Oort, R., Lindeboom, M., Koolman, X., Portrait, F.: Regional variation in public long-term home care : regional differences in the Netherlands and the role of patient experiences. Nordic J. Health Econ. 8, 87–113 (2020). https:// doi. or g/ 10. 5617/ njhe. 6670 16. Andersen, R., Newman, J.F.: Societal and individual determinants of medical care utilization in the United States. Milbank Q. 83, 1–28 (2005). https:// doi. org/ 10. 2307/ 33496 13 17. Cutler, D.M., Sheiner, L.: The geography of medicare. Am. Econ. Rev. 89, 228–233 (1999). https:// doi. org/ 10. 1257/ aer. 89.2. 228 18. Gupta, D., Biswas, D., Kabiraj, P.: COVID-19 outbreak and Urban dynamics: regional variations in India. GeoJournal 87, 2719–2737 (2022). https:// doi. org/ 10. 1007/ s1070802110394-6 19. Schulz, E.: Determinants of institutional long-term care in Germany. Enepri Research Report. 115, (2012) 20. Pickard, L.: Substitution between formal and informal care: a `natural experiment’ in social policy in Britain between 1985 and 2000. Ageing Soc. 32, 1147–1175 (2011). https:// doi. org/ 10. 1017/ s0144 686x1 10008 33 21. Alders, P., Costa-Font, J., de Klerk, M., Frank, R.: What is the impact of policy differences on nursing home utilization? The cases of Germany and the Netherlands. Health Policy 119, 814– 820 (2015). https:// doi. org/ 10. 1016/j. healt hpol. 2015. 02. 006 22. Charles, K.K., Sevak, P.: Can family caregiving substitute for nursing home care? J. Health Econ. 24, 1174–1190 (2005). https:// doi. org/ 10. 1016/j. jheal eco. 2005. 05. 001 23. Freedman, V.A.: Family structure and the risk of nursing home admission. J. Gerontol.: Ser. B 51B, S61–S69 (1996). https:// doi. org/ 10. 1093/ geronb/ 51B.2. S61 24. van Houtven, C.H., Norton, E.C.: Informal care and health care use of older adults. J. Health Econ. 23, 1159–1180 (2004). https:// doi. org/ 10. 1016/j. jheal eco. 2004. 04. 008 25. Wolf, D.A., Freedman, V., Soldo, B.J.: The division of family labor: care for elderly parents. J. Gerontol. B Psychol. Sci. Soc. Sci.Gerontol. B Psychol. Sci. Soc. Sci. 52B, 102–109 (1997). https:// doi. org/ 10. 1093/ geronb/ 52b. speci al_ issue. 102 26. Bundesamt für Bauwesen und Raumordnung: INKAR indikatoren und karten zur raumund siedlungsentwicklung, www. inkar. de, (2023) 776 A.Herr et al. 27. Kenney, G.M., Dubay, L.C.: Explaining area variation in the use of medicare home health services. Med. Care 30, 43–57 (1992). https:// doi. org/ 10. 1097/ 00005 65019920 100000004 28. McAuley, W.J., Spector, W., van Nostrand, J.: Formal home care utilization patterns by rural–urban community residence. J. Gerontol. Ser. B. 64B, 258–268 (2009). https:// doi. org/ 10. 1093/ geronb/ gbn003 29. Schmitz, H., Stroka, M.A.: Health and the double burden of fulltime work and informal care provision — evidence from administrative data. Labour Econ. 24, 305–322 (2013). https:// doi. org/ 10. 1016/j. labeco. 2013. 09. 006 30. Bakx, P., de Meijer, C., Schut, F., van Doorslaer, E.: Going formal or informal, who cares? The influence of public long-term care insurance. Health Econ. 24, 631–643 (2015). https:// doi. org/ 10. 1002/ hec. 3050 31. Eibich, P., Ziebarth, N.R.: Examining the structure of spatial health effects in Germany using hierarchical bayes models. Reg. Sci. Urban Econ. 49, 305–320 (2014). https:// doi. or g/ 10. 1016/j. regsc iurbe co. 2014. 06. 005 32. Kümpel, C.: Do financial incentives influence the hospitalization rate of nursing home residents? Evidence from Germany. Health Econ. 28, 1235–1247 (2019). https:// doi. org/ 10. 1002/ hec. 3930 33. Statistisches Bundesamt: Statistischer bericht - pflegevorausberechnung - deutschland und bundesländer - 2022 bis 2070, www. desta tis. de/ DE/ Themen/ Gesel lscha ftU mwelt/ Gesun dheit/ Pfleg e/_ inhalt. html#_ 5o9vc unzo, (2020) 34. Statistik der Bundesagentur für Arbeit: Bestand an gemeldeten arbeitsstellen nach ausgewählten berufen und wirtschaftszweigen, project-no.: [292547], statistik.arbeitsagentur.de/DE/, (2023) 35. Anselin, L.: Under the hood Issues in the specification and interpretation of spatial regression models. Agric. Econ. 27, 247–267 (2002). https:// doi. org/ 10. 1111/j. 15740862. 2002. tb001 20.x 36. LeSage, J.: Spatial econometrics. In: Handbook of research methods and applications in economic geography. Edward Elgar Publishing (2015) 37. Manski, C.F.: Economic analysis of social interactions. J. Econ. Perspect. 14, 115–136 (2000). https:// doi. org/ 10. 1257/ jep. 14.3. 115 38. Revelli, F.: Performance rating and yardstick competition in social service provision. J. Public Econ. 90, 459–475 (2006). https:// doi. org/ 10. 1016/j. jpube co. 2005. 07. 006 39. Moscone, F., Knapp, M., Tosetti, E.: Mental health expenditure in England: a spatial panel approach. J. Health Econ. 26, 842–864 (2007). https:// doi. org/ 10. 1016/j. jheal eco. 2006. 12. 008 40. Drukker, D.M., Peng, H., Prucha, I.R., Raciborski, R.: Maximum likelihood and generalized spatial two-stage least-squares estimators for a spatial-autoregressive model with spatial-autoregressive disturbances. Stand. Genomic Sci. (2013). https:// doi. or g/ 10. 1177/ 15368 67X13 01300 201 41. Elhorst, J.P.: Spatial econometrics. Springer, Berlin, Heidelberg (2014) 42. Wooldridge, J.M.: Two-way fixed effects, the two-way mundlak regression, and difference-in-differences estimators. SSRN Electron. J. (2021). https:// doi. org/ 10. 2139/ ssrn. 39063 45 43. Dale, M.R.T., Fortin, M.-J.: Spatial autocorrelation and statistical tests: Some solutions. J. Agric. Biol. Environ. Stat. 14, 188–206 (2009) 44. Sünderkamp, S., Weiß, C., Rothgang, H.: Analyse der ambulanten und stationären Pflegenoten hinsichtlich der Nützlichkeit für den Verbraucher. Pflege 27, 325–336 (2014). https:// doi. org/ 10. 1024/ 10125302/ a0003 79 45. Herr, A., Hottenrott, H.: Higher prices, higher quality? Evidence from German nursing homes. Health Policy 120, 179–189 (2016). https:// doi. org/ 10. 1016/j. healt hpol. 2016. 01. 008 46. Herr, A., Nguyen, T.-V., Schmitz, H.: Public reporting and the quality of care of German nursing homes. Health Policy 120, 1162–1170 (2016). https:// doi. org/ 10. 1016/j. healt hpol. 2016. 09. 004 Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.