Cost-effectiveness of prehabilitation of elderly frail or pre-frail patients prior to elective surgery (PRAEP-GO) versus usual care – Protocol for a health economic evaluation alongside a randomized controlled trial

In the economic analyses, all patients included in the trial will be analysed. As the outcomes may differ depending on frailty status [26], the type and indication of the planned elective surgery (e.g. heart surgery, orthopaedic surgery, surgery for oncologic indication, etc.) and demographic variables, such as age and sex, subgroup analyses will be conducted given a sufficient size of a subgroup. For the planned subgroup analyses, the study population will be stratified by frailty status into a frail and pre-frail patient group according to the inclusion criteria. The type of planned elective surgery will be defined according to the main surgical diagnoses at baseline and divided into oncologic and non-oncologic indications, and by type of surgery (e.g., heart surgery, orthopaedic surgery, etc.).

Patients in the intervention group will receive the PRAEP-GO intervention, a complex intervention comprising multiple elements: A) frailty screening, B) a shared decision-making (SDM) conference and C) a 3-week individualized multimodal prehabilitation program before elective surgery as described in Table 1.

The primary outcome from the societal perspective is the incremental quality adjusted life years (QALYs) at 12-months follow-up postoperatively. QALYs will be derived from the EQ-5D-5L questionnaire at 3-, 6-, 9-, and 12-months postoperatively and valuated by the German value set. According to the German value set for general German population, the best health state (11111) is valued with 1, death is valuated with 0 and the worst outcome (55555) is valuated with -0.661 [35]. All deaths will be identified at the end of the trial by querying the regional death registries. QALY was chosen for the health economic evaluation because a German valuation set is available [35, 43] and the use of the outcome type in the evaluation of prehabilitation is internationally recognised, as a recent systematic review has shown [19].

The primary outcome from the health care payer (German SHI) perspective is the "avoidance of worsening of the Level of Care-Dependency (LCD)”, which is a dichotomous variable derived from the ordinally scaled LCD with six levels, ranging from level 0 (no impairment of independence or no disability) to level 5 (most severe impairment with special needs for nursing care or most severe disability) measured at 12 months post-surgery and compared to the baseline.

The LCD will be measured and determined by the study personnel during a home visit using the German New Assessment Tool of Care Dependency (Neues Begutachtungsinstrument, NBI) [20]. The NBI assesses five categories: self-care, ability to self-manage the disease, ability to arrange daily life, mobility, and cognitive and communicative ability, resulting in a total maximum score between 0 and 100 points. These points are derived from the weighting of these categories, with the highest weighs assigned to self-care (40%) and ability to self-manage the disease (20%) [36]. The resulting sum of points can be translated into LCD scores: A person who is independent in all areas and not impaired in their abilities receives 0 points, while a person who experiences the greatest possible impairment in all areas of life, affecting their independence or abilities, is awarded 100 total points. A score of 12.5 points corresponds to LCD 1, 27 points to LCD 2, 47.5 points to LCD 3, 70 points to LCD 4, and 90 points to LCD 5. The requirements for LCD 5 are also met if someone has completely lost their gripping, standing, and walking functions - regardless of the score achieved in the six categories. In the context of the PRAEP-GO RCT, if a person dies, they are also assigned LCD 5.

The probability of worsening of LCD will be measured by comparing the patient status at 12 months with the baseline value: All patients experiencing deterioration in health status or independence, as measured by the NBI, i.e., those patients with a higher LCD compared to baseline, are assigned a score of 1. All other patients receive a score of 0. We decided to prioritize the deterioration of LCD as it represents the most undesirable outcome for both the patient and the payer, and one that should be prevented. LCD was selected as the primary outcome due to its considerable relevance to the German SHI perspective. For the payer, the instrument is an important measure of disability that is directly associated with long-term and short-term care expenses incurred by the SHI. From the societal perspective, the outcome will be analysed in secondary analyses.

From the health care provider perspective, incremental length of stay in the intensive care unit (ICU) measured as incremental days stayed in the ICU at discharge will be considered as primary outcome. Overall length of stay (LOS) and complications during the hospital stay will be considered as secondary outcomes. The ICU, LOS (LOS = Discharge date - Admission date +1) and complications based on secondary diagnoses will be reported by participating hospitals on the basis of billing data. The complications will be classified according to Clavien-Dindo classification [38, 39] or comprehensive complication index [40‐42].

Cost categories and types of data used for the health economic evaluation are illustrated in Fig. 2A and B and described in more detail in the Additional file 1. The type of costs covered in the analysis depends on the adopted perspective (see Fig. 2B). While medical intervention costs are calculated in the same way for all perspectives (i.e., societal, SHI, and health care provider perspectives), the costs of hospital stay for surgery are calculated in different ways depending on the perspective. For instance, from the SHI perspective, only the hospital billing dataset, which represents the costs to the SHI, will be considered. From the societal perspective, additionally investment costs will be included into the analysis; from the provider perspective, for a set of hospitals with available data, the difference between actual cost data and billing data will be considered. The resource use during the 12-months follow-up begins on the first postoperative day and will be considered from the societal and SHI perspectives. On the contrary, opportunity costs of informal care and of the participation on the SDM will only be taken into account from the societal perspective. In the German SHI, patients are required to contribute to almost all services provided [44]. User charges of patients constitute 10% of the price of medical goods, such as pharmaceutical or auxiliary medical devices, with a minimum of 5 Euros and a maximum of 10 Euros. User charges for therapeutic services, nursing care at home or intensive care at home amount to 10% of costs and 10 Euros per prescription. The user charges for inpatient care and rehabilitation amount to 10 Euros per calendar day. The described user charges incurred by patients will be considered from the societal perspective while being subtracted from costs assessed from the SHI perspective.

Measurement and valuation of resource use and costs is described in detail in the Additional file 1. Both resource use and costs for the economic evaluation will be estimated using a bottom-up approach. The approach of data collection depends on the trial phase. The prehabilitation personnel collects and measures the resource use during the PRAEP-GO intervention. The costs and resource use during the 12 months follow-up is patient-reported, collected by study personnel via a questionnaire during a home visit, or via a telephone survey. Resource use will ultimately be valued by the health economists. Costs of hospital stay will be reported by participating hospitals. The level of aggregation of the resource use depends on the health care setting where health care is provided.

All cost estimates from previous years will be converted into Euro 2022. The year 2022 was selected as it represents the year in which most patients had their index hospital stay, which likely presents the biggest cost block. Cost estimates will be adjusted for inflation using the average annual inflation rate in health care of 2.1% in 2022 in Germany [52]. Costs incurred in later years will be deflated accordingly using the annual average index reported by European Central Bank [53].

Due to the broad inclusion criteria of the pragmatic trial, i.e., different surgical indications resulting in different types of elective surgery, we anticipate that the study population may be highly imbalanced at baseline despite a well conducted random allocation of trial participants which usually should ensure a well-balanced distribution of observable and unobservable characteristics of study population between groups [55‐57]. As baseline imbalance could introduce biases [58], we decided to apply adjustment for baseline age, sex, Charlson Comorbidity Index (CCI), surgery diagnosis as a proxy for the type of surgery and surgical risk estimate based on the type of surgery, as defined by the 2022 guideline from the European Society of Cardiology (ESC) [59], if standardised mean difference (SMD) in these variables is larger than 10%. SMD (Cohen’s d [60]) is defined as difference in means of a variable (e.g., age) between two groups divided by the pooled standard deviation from the mean of the variable among the participants of the two groups. The measure ranges between +/-1 to 0, with zero indicating perfect balance. SMD has been mostly used to evaluate the balance between groups in propensity score matching studies [61], but also in RCTs [62]. The CCI was selected as adjustment variable to account for health impairment. This selection is based on the recognition that CCI may be linked to costs and outcomes independently of any intervention [63], while it is not affected by the intervention itself. To account for unobserved characteristics of hospitals where patients were recruited and where the planned surgeries were performed, a dummy variable will be created for each hospital and incorporated into the analysis. The adjustment will be achieved by including the variables in the regression model and the results will be compared to the unadjusted model [57]. Baseline characteristics of the study population will be presented in tabularly format per group. Continuous (e.g. age) data will be presented in means and standard deviation or medians and interquartile ranges. Ordinal data will be presented as medians and interquartile ranges. Categorical data (e.g. sex) will be presented as percentages.

After database closure in July 2024, available data will be assessed towards the amount of missing data, patterns of missingness and the type of missingness of data (i.e., missing completely at random, missing at random, missing not at random). As recommended by the German Institute of Quality and Efficiency in Health Care (IQWiG), if the proportion of study population not evaluated due to missing data is below 20%, the primary analyses will be conducted using complete case analysis [34]. Imputation methods will be employed in the secondary analysis. If the proportion of study population not evaluated due to missing at random data is 20% or higher, appropriate imputation methods will be employed in the primary analysis.

In both the primary and secondary analyses, the most appropriate solution for imputing missing data will be selected for each data type and outcome: Missing demographic data will be completed, where possible, using hospital billing data sets and administrative data of the BARMER SHI fund for trial participants insured with BARMER (see description of settings and location). The BARMER data will also be used for validation and completion of resource use and cost-data. If CCI is missing at baseline, the second available datapoint will be considered, as CCI is unlikely to considerably change within the observation period. If EQ-5D-5L domains, WHODAS 2.0 or NBI domains are missing at random, the missing values will be imputed using the Multivariate Imputation by Chained Equation (MICE) algorithm as suggested by [64, 65] using R packages “miceadds” and “mice”. Patterns of distribution of cost and resource data will also be evaluated and imputed by appropriate multiple imputation methods. The description of missing data and the applied methods of handling missing data will be reported as suggested in the literature [66‐69]. The outcome and cost data at baseline and 12 months postoperatively will be presented depending on the data type. Continuous data will be presented in means and standard deviation or in medians and interquartile ranges. Ordinal data will be presented as medians and interquartile ranges. Categorical data will be presented as percentages. Effects will be presented in mean differences, or in odds ratios and 95% confidence intervals.

Results of conducted sensitivity analyses will be plotted in a cost-effectiveness plane per outcome. The probability of the PRAEP-GO intervention being cost-effective at different willingness-to-pay thresholds will be visualized as cost-effectiveness acceptability curves, derived from conducted sensitivity analyses [78, 79]. The maximum willingness-to-pay threshold considered will be 100,000 Euros per QALY gained and per unit of risk reduction in LCD. If the probability of the intervention being cost-effective does not reach 1 at the threshold of 100,000 Euros, the threshold will be further increased until the probability of 1 is reached. The threshold of 100,000 Euros was chosen arbitrarily for several reasons. First, in Germany, there is currently no specific threshold defined [34]. Second, it is above cost-effectiveness thresholds valid in other countries, such as England [80] or the Netherlands [81]. Third, it was used in other health economic evaluations, for instance, in the evaluation by Fernandes (2017) [82]. All analyses will be conducted in R using RStudio.