A Model-Based Estimation of Annual Long-Term Care Costs in Germany Following Post-Operative Cognitive Dysfunction (POCD) in Elderly Patients

The long-term care insurance (“Pflegeversicherung”) in Germany is financed by income-dependent charges of the national workforce and experienced significantly increasing expenditures during the last years. From 2013 until 2019 they increased from 23.2 billion EUR to 40.7 billion EUR, which represents an increase by nearly 75% ¹. The increase in expenditures can be explained by an expansion of services which are paid by the long-term care insurance as well as the demographic development of the German population, which is characterized by an increase in the older population ². In light of potential future losses incurred by the long-term care insurance, the need for cost reductions becomes obvious. Cost implications stemming from other sectors of welfare state, e.g. the health care sector, become increasingly important in this context.

Post-operative cognitive dysfunction (POCD) is a frequently occurring complication after surgeries, not only but especially among elderly patients aged 65 years and above ^{3, 4}. It can be described as a relative decline in patients` cognitive function following surgery compared to pre-surgery and might trigger dependency on care giving in a substantial share of elderly patients. Post-operative delirium (POD) is in many cases preceding a temporary or even sustained POCD ^{5, 6, 7}. This is especially the case when several risk-factors like pre-OP dementia, or other pre-existing comorbidities are present ^{8, 9}. As a result, many patients remain dependent on care giving after the initial hospitalization either temporarily or permanently. In light of the high operation numbers in elderly patients aged 65 years and above in Germany (>6.4 mio in 2019), who often carry a number of POCD risk factors, the long-term care burden stemming from POD and subsequent POCD has not been comprehensively assessed and published to date but might be substantial.

It is worth noticing that the existing nomenclature for perioperative cognitive deficits has recently been revised and the Nomenclature Consensus Working Group has made recommendations for a refined naming of perioperative cognitive dysfunctions ¹⁰. According to their recommendation neuro-cognitive disorders (PNDs) should be defined differently depending on the time of diagnosis in the perioperative period in order to better distinguish them from neurocognitive disorders of other genesis. Preoperative neurocognitive disorders are referred as mild / major neurocognitive disorders (NCD) or pre-existing cognitive impairment (PreCI). If a neurocognitive disorder occurs during the early postoperative intra-hospital phase, it is defined as emergence delirium (e.a. agitation after first awakening) and postoperative delirium (POD). If there is delayed cognitive recovery in the in-patient and post-hospital setting until 30 days after the operation, this condition can be described as delayed neurocognitive recovery (dNCR). All postoperative neurocognitive disorders that persist for more than 30 days are referred as "mild / major NCD (postoperative)" (formerly known as POCD). To emphasize the clinical association of NCD after surgery, the term "postoperative" is used for up to one year after surgery. Outside this period suffix “postoperative” is omitted. In the present analysis, we are focusing on POD and postoperative neurocognitive disorders that persist for more than 30 days and define the latter as POCD, because we will not differentiate between mild and major NCD.

Primary aim of the BioCog consortium is the development of a prospective diagnostic tool, capable of reliably predicting POD and POCD risks of patients, in order to either individually adapt and improve post-operative care or to suspend and potentially discard the initially planned surgery. For this purpose, the BioCog consortium conducted an observational study to identify suitable biomarkers by including more than one thousand patients in Berlin, Germany and Utrecht, Netherlands aged 65 years and above, undergoing an inpatient major elective surgery with general anesthesia. Specifics of the BioCog study design are reported elsewhere ¹¹. Final study results have been submitted for publication. For the present study we used age-specific incidence rates of POCD observed in the BioCog trial, to estimate post-operative long-term care costs associated with POCD.

In anticipation of the prospective diagnostic POCD risk prediction tool we seek to determine parts of the possible market size of such a tool in terms of EUR. To estimate the health-economic value of the proposed pre-operative assessment it is essential to also consider downstream costs associated with POCD, which could potentially be avoided. A major share of these downstream costs is suspected to be incurred by the long-term care insurance in Germany due to sustained POCD of a significant proportion of patients.

Health-economic literature on POD - and POCD-related costs is scarce, especially literature on costs borne by long-term care insurances. According to published studies in the field of health-economics, it could be shown that POD significantly contributes to increased inpatient costs ^{12, 13, 14}. It has been reported that new onset of delirium during hospitalization was associated with additional inpatient costs of approximately 1300 EUR compared to non-delirium cases in Germany, with the majority of additional costs (approximately 60%) stemming from increased inpatient nursing personnel costs ¹⁴. Budget impact estimates of POD and subsequent POCD state total direct 1-year health care costs for the US between $143 billion to $152 billion ¹⁵.

Analyses of the financial burden of POCD to the German long-term insurance caused by premature dependency on care giving could not be found in the present literature. There is a lack of published, transparent, and reproducible estimates of the long-term care burden in Germany attributable to POD/POCD, especially within the new framework of “Pflegegrade” – a formal system which assigns a patient to one out of five categories depending on the degree of long-term care intensity. The BioCog group therefore decided to develop a health-economic model estimating the cost-consequences attributable to POCD for the German long-term care insurance, based on the BioCog trial results and other published data.

The results of the base case are displayed in Table 5. Case numbers were calculated as described above and yield a total of 288,375 POCD cases annually. Applying monetary costs of the long-term care insurance in the different types of long-term care schemes described above, yields annual costs triggered by POCD of approximately 1.6 billion EUR in the base case of the model. The relative distribution of POCD related costs across long-term care schemes is shown in Figure 4. The highest share of costs (46%, or approximately 60mio EUR) is devoted to long-term care money (Pflegegeld) as compensation for relatives providing care. Care provided by professionals in an outpatient setting accounts for 38% and inpatient care for 15% of calculated additional costs.

Figure 4.Relative distribution of POCD-related costs across nursing schemes

Several deterministic sensitivity analyses were performed, and the results are shown in a tornado diagram (Figure 5) to investigate the uncertainty surrounding the model results. We in- and decreased the age-specific incidence rates and the average amount of OPS-codes per patient by +/-10% and recalculated the relevant POCD cases and additional costs with the aim to evaluate the effect of uncertain and varying POCD case numbers. In addition, we investigated the effect of assuming half of the POCD cases transitioning from long-term care intensity 0 to 1 and from intensity 0 to 3 instead of assuming all POCD cases transitioning from intensity 0 to 2.

Figure 5.Results of deterministic sensitivity analyses (Tornado Diagram)

Regarding the three chosen parameters and assumptions, it becomes obvious that the highest share of uncertainty surrounding the model results stems from the assumption of all POCD cases transitioning from long-term care intensity 0 to 2. Assuming half of the POCD cases transitioning from care intensity 0 to 1 and the other half from 0 to 2 would decrease annual costs to the long-term-care insurance to approximately 796mio EUR. On the other hand, assuming half the POCD cases transitioning from care intensity 0 to 3 and the other half again from 0 to 2 would increase total costs to 2.3 billion EUR.

While a 10% in- or decrease in POCD incidence rates directly translates into a 10% in- or decrease in total long-term care costs, a corresponding in- and decrease of the average amount of chapter 5 OPS-codes per patient and inpatient surgery would result in annual costs of either 1.4 billion EUR (10% increase) or 1.7 billion EUR (10% decrease). The diverging effect compared to an alteration of POCD incidences stems from the distribution of cases in respective age-groups and the fact, that OPS numbers are divided by this factor.

A strength of the present analysis lies in its transparency and the utilization of publicly available data, together with recently established POCD incidence rates in Germany and reasonable assumptions. However, the results of the sensitivity analyses vary – in one case even significantly – indicating the degree of uncertainty of assumptions and inputs that were used.

The reported results of our model-based calculations depend on the applied incidence rates of POCD. Therefore, significant effort was made to obtain reliable estimates of age-specific incidence rates across surgery types from the BioCog study. The fact that we applied incidence rates across surgery types introduces uncertainty of the modelled results, because the data base used for estimating POCD patient numbers (OPS 2019 statistic) reflects slightly different shares of operations with regards to the surgery site compared to the BioCog sample from which POCD incidence was inferred. This is important to note, since most publications on POCD incidences focus on specific types of surgeries and report highly heterogeneous numbers ^{19, 20, 21, 22}.

For the base case, we used a controlled definition of POCD based on a reliable change index as suggested by Rasmussen et al., 2001. This definition results in lower POCD incidence compared to more liberal definitions. To give an example: A recent publication on recommendations for the nomenclature of cognitive change associated with anesthesia and surgery suggests to report postoperative cognitive dysfunction more in line with the DSM-5 diagnosis of Neurocognitive disorder (mild/major NCD) and use the former ‘POCD’ as a specifier to indicate its coincidence with surgery for up to one year postoperatively ¹⁰. Mild NCD with a more liberal cut-off of 1-2 standard deviations would result in higher POCD case numbers and subsequently in higher costs to long-term care insurance.

It should also be noted, that the BioCog sample – from which we drew the applied POCD incidence rates – represents a younger cohort than the chapter 5 OPS utilization sample that was used to estimate the number of patients undergoing surgery in Germany in 2019. Since POCD-incidence is highly dependent on patient age, the model rather underestimates the true number of POCD cases.

In addition to that, POCD incidence rates obtained from the BioCog study might be influenced by the trial`s drop-out rate: POCD data at 3 months is available for 641 out of 933 patients. Low follow-up rates (69% for cognitive testing in BioCog) are common in POCD trials, e.g. the ISPOCD study reported a similar low POCD prevalence (10%) with a comparable follow-up rate of 78% ⁴. This comparably high number of patients lost-to-follow-up was anticipated at the outset of the BioCog study and detailed analyses of patients lost to follow-up were performed. In our context, an increased drop-out rate is considered to mask actual POCD rates, rendering our applied POCD rates as conservative approximations of the reality.

A limitation of the present study involves the fact, that cost consequences of transitions from long-term care intensity 0 to 2 (Pflegegrad 2) were modelled. We investigate the effect of exemplary alternative transition distributions in corresponding sensitivity analyses, but in fact there are many other possible transitions (e.g. from care intensity 1 to 4, from 2 to 5 etc. and even switches between long-term care types) that were not modelled but each triggering different cost consequences depending on the shares of patients in each segment of inpatient or outpatient long-term care. These aspects were not modelled, because there is a lack of evidence that could serve as a base for reasonable assumptions and guide our calculations. The fact that we only modelled one possible switch from long-term care intensity 0 to 2 – but in fact there many more – indicates that the long-term care costs attributable to POCD are likely to be even higher than we report. Future analyses should focus on determining the exact amounts and shares of patients transitioning from one care segment to another.

In light of our conservatively chosen incidence rates, the analyzed chapter 5 OPS utilization numbers as a representative base for POCD case number calculations, actually reflect performed operations and not cases or patients. We sought to control for that aspect by excluding supplemental OPS-codes not reflecting separate surgeries, and by dividing the resulting number of chapter 5 OPS-codes by the average number of chapter 5 OPS-codes per patient observed in a sub-sample of the BioCog study. Dividing chapter 5 OPS utilization by the average number of chapter 5 codes per patient introduces additional uncertainty, because it is unclear whether this factor is the same in all German patients undergoing surgery captured in the OPS data. Therefore, we investigated the effect of varying average numbers of chapter 5 OPS-codes per patient in corresponding sensitivity analyses.

Currently, a large multicenter clinical trial is recruiting patients in Germany who will undergo a multimodal and multidisciplinary intervention after surgery with the aim to reduce POD rates by at least 40% and to reduce POCD rates by at least 20% ²³. Assuming a 20% decrease in POCD rates within the framework of the presented model, would translate into a reduction of POCD related costs of around 312mio EUR in the long-term care insurance according to the base case model. Additional savings in the health care sector are likely due to reduced expenses for diagnostic procedures associated with the staging of POCD patients for different long-term care intensities. Given the assumed effectiveness of the multidisciplinary intervention, a prospective diagnostic tool that could reliably detect POCD before surgery, would have a significant value from a health- and socio-economic perspective. The precise value of such a tool in terms of EUR would depend on the treatment consequence of a detected suspected POCD case and associated costs of these consequences. Test-performance in terms of sensitivity and specificity as well as associated costs of applying the test on a patient are additional determinants of the health- and socio-economic value. In the present context, treatment consequences are limited to:

cancellation of surgery, if justifiable from a medical perspective

suspension of surgery, if improvement of test-outcome at a later point of time is likely

modification of treatment path, e.g. application of multimodal and multidisciplinary interventions, modified anesthesia or other measures

These consequences would need to be precisely investigated with regards to costs and effectiveness for an exact estimation of the health- and socio-economic value of a prognostic POCD detection tool.

A first and transparent model-based analysis of POCD associated costs in the German social long-term care insurance could be developed and evaluated. The analysis reveals significant long-term care costs associated with POCD of approximately 1.6 billion EUR annually. Several assumptions and features of the cost-model ensure conservative estimates of the results, suggesting that the actual financial burden of POCD to the long-term care insurance is even higher. For a precise determination of POCD related long-term care costs more research needs to be devoted to aspects surrounding the consequences of POCD with respect to the long-term care insurance. These aspects include age-specific POCD incidence rates, numbers and shares of patients transitioning from no dependency on long-term care to other segments of long-term care intensity and type. Additional studies should further reveal whether shares of patients are altered between different long-term care schemes due to the development of POCD, and consider effects on expenses for technical long-term care aids (Pflegehilfsmittel), support for improvements in home environments (Wohnumfeld-Verbesserungsmaßnahmen), transfer payments to pension funds, unemployment insurance, statutory health insurances and other legal financial support of the long-term care insurance.

The research leading to these results has received funding from European Union funded Seventh Framework Research Program ^{FP7/2007–2013}, under grant agreement No. HEALTH-F2-2014-60246, BioCog (Biomarker Development for Postoperative Cognitive Impairment in the Elderly), www.biocog.eu.

All authors contributed to the study conception and design. The first draft of the manuscript was written by Simon A. Weber and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Simon A. Weber¹, Malte Pietzsch¹, Claudia Spies², Fatima Yürek², Daniel Hadzidiakos², Friedrich Borchers², Florian Lammers-Lietz², Sophie K. Piper^3,5, Jochen Kruppa^3,5, Georg Winterer^2,4, on behalf of the BioCog consortium

Cellogic GmbH, Niedstrasse 21, 12159 Berlin

Department of Anesthesiology and Intensive Care Medicine (CCM, CVK), Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Augustenburger Platz 1, 13353 Berlin, Germany.

Institute of Biometry and Clinical Epidemiology, Charité - Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt-Universität zu Berlin, and Berlin Institute of Health, Charitéplatz 1, 10117 Berlin, Germany

Pharmaimage Biomarker Solutions GmbH, Biotech Park Berlin-Buch, Robert-Rössle-Str. 10, 13125 Berlin, Germany

Berlin Institute of Health (BIH), Anna-Louisa-Karsch-Strasse 2, 10178 Berlin, Germany

1. 1.Bundesministerium für Gesundheit, B. (2020) . Zahlen und Fakten zur Pflegeversicherung
   Search at Google Scholar

1. 2.Arentz C. (2019) Szenarien zur zukünftigen Finanzentwicklung der Sozialen Pflegeversicherung.Wissenschaftliches Institut der PKV.
   Search at Google Scholar

1. 3.Gold S, Forryan S. (2019) Postoperative cognitive decline: a current problem with a difficult future.Trends in Anaesthesia and Critical Care. 24, 49-58.
   Search at Google Scholar

1. 4.Moller J, Cluitmans P, Rasmussen L, Houx P, Rasmussen H. (1998) Long-term postoperative cognitive dysfunction in the elderly:. ISPOCD1 study.The Lancet 351(9106), 857-861.
   Search at Google Scholar

1. 5.Rudolph J, E R Marcantonio, D J Culley, Silverstein J, Rasmussen L. (2008) Delirium is associated with early postoperative cognitive dysfunction.Anaesthesia. 63(9), 941-947.
   PubMed·View article·Search at Google Scholar

1. 6.T G Monk, C. (2011) Postoperative cognitive disorders.Curr Opin Crit Care. 17(4), 376-81.
   Search at Google Scholar

1. 7.Rundshagen I. (2014) Postoperative cognitive dysfunction.Dtsch. , Arztebl Int 111(8), 119-25.
   PubMed·View article·Search at Google Scholar

1. 8.T G Fong, Davis D, M E Growdon, Albuquerque A, S K Inouye. (2015) The interface between delirium and dementia in elderly adults.Lancet. , Neurol 14(8), 823-832.
   Search at Google Scholar

1. 9.T N Robinson, C D Raeburn, Z V Tran, E M Angles, L A Brenner. (2009) Postoperative delirium in the elderly: risk factors and outcomes.Annals of surgery. 249(1), 173-178.
   Search at Google Scholar

1. 10.Evered L, Silbert B, D S Knopman, D A Scott, S T DeKosky. (2018) Recommendations for the nomenclature of cognitive change associated with anaesthesia and surgery-2018.Br. 121(5), 1005-1012.
   PubMed·View article·Search at Google Scholar

1. 11.Winterer G, Androsova G, Bender O, Boraschi D, Borchers F. (2018) Personalized risk prediction of postoperative cognitive impairment - rationale for the EU-funded BioCog project.Eur Psychiatry. 50, 34-39.
   PubMed·View article·Search at Google Scholar

1. 12.D L Leslie, E R Marcantonio, Zhang Y, Leo-Summers L, S K Inouye. (2008) One-year health care costs associated with delirium in the elderly population.Arch Intern Med. 168(1), 27-32.
   Search at Google Scholar

1. 13.Brown C H t, Laflam A, Max L, Lymar D, K J Neufeld. (2016) . The Impact of Delirium After Cardiac Surgical Procedures on Postoperative Resource Use.Ann Thorac Surg 101(5), 1663-9.
   Search at Google Scholar

1. 14.Weinrebe W, Johannsdottir E, Karaman M, Fusgen I. (2016) What does delirium cost? An economic evaluation of hyperactive delirium.Z. , Gerontol Geriatr 49(1), 52-8.
   PubMed·View article·Search at Google Scholar

1. 15.D L Leslie, S K Inouye. (2011) The importance of delirium: economic and societal costs.J. , Am Geriatr Soc 59, 241-3.
   Search at Google Scholar

1. 16.Bundesamt Statistisches, D. (2020) Fallpauschalenbezogene Krankenhausstatistik (DRG-Statistik). Operationen und Prozeduren der vollstationären Patientinnen und Patienten bis zum kodierbaren Endpunkt .
   Search at Google Scholar

1. 17.Rasmussen L, Larsen K, Houx P, Skovgaard L, Hanning C. (2001) The assessment of postoperative cognitive function.Acta Anaesthesiologica Scandinavica. 45(3), 275-289.
   Search at Google Scholar

1. 18.Bundesamt Statistisches, D. (2020) Pflegestatistik - Pflege im Rahmen der Pflegeversicherung. , Deutschlandergebnisse
   Search at Google Scholar

1. 19.M J Needham, C E Webb, D C Bryden. (2017) Postoperative cognitive dysfunction and dementia: what we need to know and do.Br. 119, 115-125.
   Search at Google Scholar

1. 20.Holmgaard F, A G Vedel, L S Rasmussen, O B Paulson, J C Nilsson. (2019) The association between postoperative cognitive dysfunction and cerebral oximetry during cardiac surgery: a secondary analysis of a randomised trial.Br. , J Anaesth 123(2), 196-205.
   PubMed·View article·Search at Google Scholar

1. 21.W X Li, R Y Luo, Chen C, Li X, J S Ao. (2019) Effects of propofol, dexmedetomidine, and midazolam on postoperative cognitive dysfunction in elderly patients: a randomized controlled preliminary trial.Chin Med. 132(4), 437-445.
   Search at Google Scholar

1. 22.Tzimas P, Samara E, Petrou A, Korompilias A, Chalkias A. (2018) The influence of anesthetic techniques on postoperative cognitive function in elderly patients undergoing hip fracture surgery: General vs spinal anesthesia.Injury. 49(12), 2221-2226.
   Search at Google Scholar

1. 23.Sanchez A, Thomas C, Deeken F, Wagner S, Kloppel S. (2019) Patient safety, cost-effectiveness, and quality of life: reduction of delirium risk and postoperative cognitive dysfunction after elective procedures in older adults-study protocol for a stepped-wedge cluster randomized trial. , (PAWEL Study).Trials 20(1), 71.
   Search at Google Scholar

1. 1.Fislage Marinus, Feinkohl Insa, Borchers Friedrich, Pischon Tobias, Spies Claudia D., et al, 2023, Preoperative thalamus volume is not associated with preoperative cognitive impairment (preCI) or postoperative cognitive dysfunction (POCD), Scientific Reports, 13(1), 10.1038/s41598-023-38673-x