FormalPara Key Points for Decision Makers

Applying micro-cost accounting in healthcare identifies improvement areas in the delivery of care or activities that are potent for substitution of care.

Fuzzy logic time-driven activity-based costing (FL-TDABC) is a time-efficient method that provides more accurate time estimates of the daily clinical practice wherein patient complexity is transparently reflected.

The largest expense within the first year of the Rheumatoid Arthritis (RA) patient journey occurs during the diagnostic phase where consultation with a rheumatologist is the main cost driver along with pharmaceutical expenditures.

Mainly depending on the severity of the illness in patients, maximum RA cycle costs (€1684) were some 22% higher than minimum costs (€1317) observed through FL-TDABC.

1 Introduction

Due to a lack of transparency in healthcare expenditures, governments and decision makers are inhibited from adequately containing the growth of costs [1]. Over the years, various management tools have been proposed as possible solutions for the high level and rapidly increasing expenses. However, strategies such as lean and six sigma cannot hold healthcare practitioners appropriately accountable for the quality and costs of care delivery due to the lack of time and resources to implement those tools [2]. Another drawback of lean and six sigma is the one-dimensional focus on processes, instead of the preferred holistic perspective [3].

Policy makers should make informed decisions with respect to the allocation of resources within healthcare, which requires transparency of the costs associated with the treatments provided [4]. Cost information can offer managerial value if costs can be linked to the factors that drive them, such as individual procedures in a care cycle. Several cost allocation methods have been developed to provide insight into the drivers of healthcare costs on a micro level [5].

The practice of cost accounting and allocation in healthcare has evolved over time, leading to, inter alia, the development of time-driven activity-based costing (TDABC). TDABC was introduced by the Harvard Business School in 2003, as an elaboration of the traditional activity-based costing (ABC) approach [6]. The TDABC methodology makes use of two variables, namely the required time commitment by a resource and the capacity cost rate (CCR) of that resource [6]. From a value-based healthcare (VBHC) perspective, calculating the costs of care delivery, i.e., the activities performed regarding the treatment of a disease, is an essential component [7]. Therefore, TDABC has been proposed as the principal cost accounting method within the VBHC framework [7].

A drawback of traditional ABC, which underlies the current Dutch reimbursement system within healthcare, is the fact that the time spent by medical staff to perform certain treatments is based on point estimates. In the case of the Dutch healthcare reimbursement system, these point estimates are not disclosed (black box): healthcare cost reimbursements comprise a lump sum for every standardized diagnosis treatment combination (DTC), of which there are around 4000. However, information about the time factor (time spent per resource) is of great importance to achieve transparency in costs, as patients experience (multi)morbid conditions that lead to variations in the duration of consultations with healthcare professionals. Especially in an era of shared decision making, time spent by medical staff is a key driver of healthcare costs. As TDABC also uses point estimates for the time required to perform medical procedures, TDABC can also be assumed to be biased [8]. To deal with the variation in time estimates regarding healthcare activities, this study combines TDABC with the fuzzy logic (FL) theory, which uses pooled time estimates instead of point estimates. The use of FL is based on the observation that medical and healthcare data tends to be subjective (fuzzy). FL-TDABC has shown to be an effective strategy in healthcare and other sectors as the full spectrum of possible conditions from a best- to worst-case scenario can be taken into account [8,9,10]. Furthermore, implementing and maintaining a TDABC model is a labor-intensive and time-consuming procedure, predominantly carried out by researchers [11].

The aim of this study was to apply the FL-TDABC methodology to the Rheumatoid Arthritis (RA) standardized care cycle in Maasstad Hospital, to estimate the treatment costs per patient, considering the variation in time spent by medical staff to perform activities within that care cycle. Variation in the actual time committed by the medical staff per patient is caused by the heterogeneity of the RA patient population in terms of, inter alia, the complexity of their medical status, i.e., experiencing multimorbidities. FL-TDABC is considered to be a less demanding cost allocation method than traditional ABC or TDABC for the determination of the time required to perform healthcare activities. Furthermore, FL-TDABC can provide information regarding the cost drivers of the standardized RA treatment cycle that can be used in the decision making concerning healthcare resource utilization and allocation. The current healthcare cost reimbursement system in The Netherlands, which is based on the ABC methodology, does not provide that information, as the time estimates underlying the calculations have not been disclosed.

2 Methods

2.1 Study Design and Population

For this TDABC study, the standardized care path for RA patients, developed at the Rheumatology and Clinical Immunology Department of Maasstad Hospital, a top clinical teaching and research hospital in Rotterdam, The Netherlands, was used [12]. For valuation purposes, 2018 cost estimates were used since 2020 and 2021 prices were affected by the coronavirus disease 2019 (COVID-19) pandemic and therefore did not represent the actual, competitive price. Identification of the patient’s medical condition corresponds with step one of the TDABC framework [13]. RA is a chronic disease causing inflammation of joints and tissues [14]. RA patients require lifelong treatment with the goal of diminishing the disease activity in patients, mainly achieved by administering disease-modifying anti-rheumatic drugs. Future treatment within the RA patient population will focus on the cost-effective and patient-centered delivery of care [15].

2.2 Data Collection

To obtain the research data, a mixture of quantitative and qualitative methodologies were applied. Quantitative data were collected via hospital reports concerning financial and human resource data. With respect to the financial data, the annual salaries of the medical staff and the direct and indirect costs of the performed activities were assessed [16]. Human resource data concerned the medical staff capacity, in full-time equivalents (FTEs) and average work days per week. Data concerning time estimates were acquired by interviewing the staff, and care activities were obtained from electronic health records in 2021. A total of six counter employees, five rheumatologists, four doctors’ assistants, three physician assistants (or physician associate)/nurse practitioners, and three rheumatology nurses (i.e. nurses specialized in rheumatology) participated in the questionnaires. A physician assistant is a nurse or paramedic who has completed advanced training and who can independently perform complex medical tasks and actions, otherwise performed by a rheumatologist. As part of the ongoing JOINT Evaluation study, evaluation of the (clinical) data reported at the Rheumatology Department is allowed by the hospital board (MEC number: T2016-76).

2.3 Fuzzy Logic Time-Driven Activity-Based Costing (FL-TDABC)

The seven-step approach established by Kaplan and Porter [13] was followed to systematically gain insight into the costs of the standardized care cycle of RA patients (Fig. 1) [13]. The first three steps served as the basis for the present study. In brief, based on the medical condition defined (step 1), the care delivery value chain (CDVC) (step 2) and detailed process map for RA patients (step 3) receiving treatment in Maasstad Hospital were determined [12]. The emphasis of this study was to investigate and elaborate on steps four to seven of the approach, as shown below (Fig. 1).

Fig. 1
figure 1

Time-driven activity-based costing framework in healthcare

Full size image

The TDABC methodology was modified by including FL in the approach (Fig. 1) [13]. Cost prices were calculated by allocation of resources to cost centers using two variables: the CCR of a resource and the time commitment by that resource [17]. The CCR is the total of all costs required to supply a resource, such as a consultation or a laboratory analysis, divided by the time required to supply that resource by the employees [17]. Since RA is a chronic disease, the duration of the care cycle was set at one year as recommended by the Harvard Business School [13]. The healthcare providers involved or other resources required and the place of care delivery were denoted per activity (i.e. procedure) that patients undergo [12].

2.4 Time Estimation of the Activities in the Process Map (Step 4)

2.4.1 Fuzzy Logic

The FL theory was developed by Zadeh to improve the logic of fuzzy data sets, i.e., data sets with imprecise information [8, 18]. Imprecise information in healthcare is predominantly the result of the variability of the time spent on procedures. FL allows managers and policy makers to make decisions based on time estimates that are closer to clinical practice, which has a positive effect on the organization of care. In this study, the methodology of the triangular fuzzy numbers (TFNs) is applied, where TFNs represent a category of fuzzy numbers reducing the ambiguity of outcomes, specifically with respect to the variation in time commitment required by the medical procedures within the diagnosis-related groups (DRGs) between caregivers [8]. TFNs are composed of three parameters (Eq. 1): the smallest possible value (a), the most promising value (b), and the largest possible value (c) [8]. The most promising value is defined as the modal value or the most frequently occurring value. Equation 1 shows the function underlying the FL estimate, which is applied to calculate the average costs of the activities (x*). As opposed to TDABC, FL accounts for the variation in possible outcomes due to the fact that the calculation is based on three outcome parameters. Within TDABC, the labor-intensive stopwatch timing is predominantly used to obtain time stamps of procedures, which is a more subjective and time-consuming method.

Equation 1 Center of gravity—fuzzy logic estimate

$${x}^{*}=\frac{\left(a +b +c \right)}{3}.$$
(1)

In TDABC, for generic, low-cost and brief procedures, standard times were used. As standard times were unavailable for X-rays of the hands, feet and lungs, as well as for laboratory tests, the time to perform imaging and laboratory tests could therefore not be included in the cost calculations. For the complex activities within the RA care cycle, time estimates were obtained via interviews or direct observations. In the FL-TDABC analysis, the time spent on complex and expensive procedures was estimated according to the FL method. Through questionnaires, rheumatology staff were asked to estimate the time they spent on activities such as consultations, patient registration and disease activity score (DAS) assessments. The smallest possible value, the most promising value, and the largest possible value in terms of the time required to perform these activities were assessed. The most promising value indicates the most frequently occurring time estimate as a result of inquiry among the medical staff.

2.5 Cost Estimation of Supplying Patient Care Resources (Step 5)

Estimates of the cost of supplying patient care resources were obtained through assessment of the direct and indirect costs related to the treatment of the medical condition RA [13].. Direct costs included all activities immediately associated with the care delivery, such as consultations, materials, laboratory processes and medication. Indirect costs consisted of general costs such as IT, housing and electricity, and these costs were allocated by the finance department via allocation keys. Honorarium costs were separated as these costs are used to calculate the FL-TDABC estimates. Laboratory costs consisted of the direct, indirect and honorarium costs of the frequently conducted blood tests, with exception of the laboratory costs in the diagnosis stage. Direct, indirect and honorarium costs concerning the x-rays were based on standard prices. Pharmaceutical costs were allocated to the drug verification procedure comprising the pharmacy assistant. Medication use by patients is accounted for in the calculation of the pharmacy costs. These direct, indirect and honorarium costs, if available, were allocated to the activities in the process map.

2.6 Estimation of the Capacity of Patient-Specific Resources and Calculation of the Capacity Cost Rate (Step 6)

The CCR was calculated by dividing the total costs per patient-specific resource by the practical capacity available. Practical CCRs were expressed as the cost per minute for the deployment of medical staff members. The practical capacity was derived by extracting the actual care delivery hours, excluding time spent on meetings, education, administration, holidays, sick leave, and breaks. Annual salary costs, including an employer mark-up of 30% (for social security and pension premiums), and the FTE per function were considered in calculating the costs per minute of the medical staff.

2.7 Total Cost Calculation of the Rheumatoid Arthritis Care Cycle (Step 7)

The final step of the TDABC model for healthcare is calculating the costs of the cycle of care for a patient [13]. The time spent per resource was multiplied by the capacity cost rate. Costs were aggregated to obtain the costs of the care delivery cycle for RA patients. The CCR per resource multiplied by the duration of the activity (in minutes) per phase, gives the costs of every process step [13]. The FL-TDABC total costs of treating a RA patient was calculated as the unweighted average of the total costs comprising all process steps for the smallest possible, most promising and largest possible values [13]. To analyze the variability in costs, the smallest, most promising and largest values are also displayed in the Results section.

2.8 Time-Driven Activity-Based Costing

To compare the cost estimates with the ABC and TDABC estimates, in TDABC, standard times are used for the TDABC method instead of the FL time inputs. Calculation of the TDABC value is based on the methodology as defined by Kaplan and Porter [13]. The CCR and the time commitment were applied to calculate the prices per activity or resource with respect to the honorarium costs [17]. Direct and indirect costs do not vary between the FL-TDABC and TDABC, apart from the honorarium costs. Therefore, instead of applying FL time estimates as in FL-TDABC, standard time point estimates were considered in TDABC. Standard times are nationally determined with respect to consultations with rheumatologists [19]. The standard time for the first consultation of an RA patient is 50 min, the consultations within 120 days after diagnosis are set at 40 min, and follow-up consultations are given 15 min. Consultations with physician assistants and nurse practitioners are set at 15 min in Maasstad Hospital. No standard times are available for other staff members (e.g. rheumatology nurses, doctor’s assistants and counter employees) and therefore the mean value of the FL estimate was applied as a reference to estimate the TDABC care cycle costs.

2.9 Activity-Based Costing (ABC)

Comparisons between the different cost accounting methods were also made with respect to the traditional ABC method. ABC costs are estimated by calculating the average DTC, in Dutch (Diagnose Behandeling Combinaties [DBC]), costs per RA patient in the first year of RA treatment. The ABC value was calculated by dividing the sum of the DTC costs by the number of patients in the RA patient population (2837 patients were included in the cost calculation). Hence, this calculation is not based on the standardized care path. The Dutch DTC codes comprise the registration, diagnosis, treatment and billing codes with respect to diseases treated in specialized medical care [20]. Cost reimbursements based on DTC codes are calculated by a grouper, an automated cost allocation system, centrally organized in The Netherlands with mandatory validation for reimbursement (see Appendix) [20]. Based on the nature and number of care activities provided to the patient, the most appropriate DTC product is selected by the system and a price calculated and communicated to the health insurer(s) [20]. In The Netherlands, over 4000 DTC codes exist for the specialized medical care provided in hospitals [20].

2.10 Statistical Analysis

To test the significance of the difference between the current ABC (i.e. DTC) model versus the FL-TDABC model and the TDABC model using standard times, two one sample t-tests were conducted. The two values concerning FL-TDABC and TDABC, obtained from the standardized, average care path and therefore considered as average total costs, were compared with the mean costs of the ABC (i.e. DTC) estimate as mentioned in the previous paragraph. The FL-TDABC and TDABC values are compared with the ABC (i.e. DTC) value, since the ABC method underlies the current costing application in the Dutch healthcare system. Because samples sizes were large and therefore the central limit theorem holds, parametric tests were conducted. Stata/SE 15 for Windows (StataCorp LLC, College Station, TX, USA) was used to perform the statistical analysis and an alpha level of 5% was considered as statistically significant.

3 Results

3.1 Study Population

The study results were based on the mapped care path of RA patients diagnosed by a rheumatologist in the Maasstad Hospital, Rotterdam, within the CDVC [12]. More than 2800 patients received medical care for RA in Maasstad Hospital. The majority of the patient population was female (71.5%), and on average, patients were 57.7 years of age (standard deviation [SD] 15.0) when first diagnosed with RA by a rheumatologist.

3.2 Time Estimates

The CDVC and detailed process map presented in the Methods section were applied as the starting point for this study [12]. Time estimates per activity and resource are shown in Table 1. The FL methodology was adopted with respect to the complex and more time-dependent healthcare activities, whereas for the remaining activities, standard time estimates were used. In absolute terms, the difference between the smallest and largest possible value for the time estimates (FL time) is the highest for the first consultation (Δ = 22 min, range 19–41) and diagnosis consultation (Δ = 15 min, range 10–25) with the rheumatologist and the education (e.g., instruction) activity of the rheumatology nurse (Δ = 15 min, range 15–30). In relative terms, appointment scheduling by counter employees (82%) and the DAS assessment/PROMs activity of the doctor’s assistant (67%) show the largest variances in the FL estimates.

Table 1 Fuzzy logic time estimations for the rheumatoid arthritis care cycle
Full size table

3.3 Costs of the Care Cycle

The calculated capacity cost rates for the (medical) staff members are depicted in Table 2, which are used for calculating the total care costs. Regarding the first year of RA treatment, rheumatologists have the highest CCR (€2.89 per minute), followed by the rheumatology nurse (€1.41) and the physician assistant/nurse practitioner (€1.24). Costs of a counter employee, doctors assistant and pharmacy assistant were less than €1 per minute.

Table 2 Capacity cost rates per staff member
Full size table

Given the CCRs, costs are calculated per CDVC step (Table 3). The total cost for treating a RA patient during the first year following diagnosis is calculated at €1497 using the FL-TDABC methodology. The diagnostics phase represents 32% of the total costs. Imaging (e.g. x-rays of the hands, feet and lungs), laboratory testing and the extensive time required for the first and second consultation with the rheumatologist account for the majority of the costs in the diagnostics phase of the CDVC. Within the CDVC, recovery and intervention account for nearly half of the costs due to the high frequency of follow-up visits within the first year, predominantly consultations with the more expensive rheumatologists and rheumatology nurses. Monitoring and managing and preparing represent 17% and 5% of the standardized care cycle costs, respectively.

Table 3 Fuzzy logic time-driven activity-based costing estimations of costs of the RA care cycle per care delivery value chain step
Full size table

The costs related to both (medical) staff and other resources are shown in Fig. 2. Consultations with rheumatologists account for two-fifths of the total costs. In chronic care settings such as rheumatology, patients visit healthcare facilities often and therefore the contribution of consultations to the total costs is significant. In particular, this holds true for the first year of the RA standardized care cycle since patients have to visit the hospital frequently for consultations and medical tests. Almost one-third of the costs are spent on diagnostics and pharmacy, including drug costs. Over 60% of the CDVC recovering phase costs are due to pharmaceutical and diagnostic costs. Since patients generally switch to higher priced biologicals in case of ineffective treat-to-target after 6 months, pharmacy expenditures are relatively high. The physician assistant and nurse practitioner generate over 15% of the costs and the rheumatology nurse 8%. The care delivery costs spent on the counter and doctor’s assistant account for less than 5% per cost driver.

Fig. 2
figure 2

Costs per provider or resource

Full size image

3.3.1 FL-TDABC versus TDABC (Standard Times) and ABC

Based on the FL-TDABC approach, given the minimum, most promising and maximum time estimates, annual costs per RA patient range from a minimum of €1317 via a most promising value of €1490 to a maximum of €1684. This outcome reflects the real-life variation in patients’ treatment regarding case-mix and multimorbidity, with a 22% difference between the minimum and maximum annual costs of RA treatment. The costs of the standardized RA care cycle totaled €1609 per patient based on TDABC and €1604 (SD €2139) per patient following the ABC methodology (i.e. DTC method). The variance with the average FL-TDABC costs of €1497 is explained by the difference in time commitment by the rheumatologist (consults) in the diagnosing and intervention phases of the CDVC. FL-TDABC uses an average of 57 min consulting time by the rheumatologist, whereas the standard time is 130 min in those phases.

3.4 Statistical Analysis

The one-sample t test showed a significant difference in costs between the FL-TDABC and ABC (DTC) methodology at a 5% alpha level (p = 0.0101, t = 2.566). Thus, the findings of the one sample t test of FL-TDABC versus ABC, the cost accounting methodology currently applied in Dutch healthcare, suggests there is a significant difference in cost estimates. Concerning the difference between the TDABC and ABC (i.e. DTC) cost accounting methods, no significant difference was found on a 5% alpha level (p = 0.895, t = − 0.131).

4 Discussion

Since TDABC estimates for care cycles are generally cumbersome to generate (through time studies) and do not allow for the variation in time spent on patients by healthcare providers, FL-TDABC is suggested as a viable alternative for cost allocation in healthcare. FL-TDABC highlights the time spent and costs related to the practice variation observed, through a relatively simple and quick querying of the relevant medical staff. The FL-TDABC model applied to the standardized RA care cycle at Maasstad Hospital shows that the diagnosing phase accounts for the bulk of the costs (32%), followed by the recovering and intervening phases (45.3% combined). Major cost drivers are consultations, more specifically with the rheumatologist, and pharmacy costs (i.e. medication). Maximum RA cycle costs for the more severe or multimorbid RA patients are 22% higher than the minimum costs, representing the less complex patients with shorter consultation times, observed using the FL-TDABC method.

TDABC based on standard times resulted in higher costs compared with the FL approach. Higher costs of the diagnosing phase in the TDABC model, largely explain the observed discrepancy between the models. As mentioned in the Methods section, ABC costs are derived from pricing DTCs. Total unadjusted TDABC cost estimates proved to be the highest, although the difference with the ABC (DTC) estimates are minimal. The FL-TDABC methodology allows for a more precise cost estimate, as the deterministic time estimates are pooled to an average and corrected for the subjectivity of variation in point estimations [21]. Furthermore, time estimates are obtained from healthcare providers who have first-hand knowledge of the patient population in terms of the required care. This leads to a better representation of the clinical daily practice that is incorporated in a value-based accounting method.

Enhancing the TDABC methodology with FL allows for adjusting the subjectivity of time estimates [8]. Furthermore, the time-consuming process of observing activities and obtaining time estimates is avoided [8]. Time estimates per healthcare activity are based on the experience of multiple staff members, and through calculation of the fuzzy numbers a more objective cost estimation can be obtained. For the traditional ABC approach underlying the Dutch DTC system, the time spent and the cost price per practitioner are not disclosed. Therefore, the ABC approach is experienced as a black box, as DTC prices are not explicitly linked to the cost drivers of healthcare [22]. From an operational perspective, managing on the basis of activities is therefore more intuitive than on the DTC codes. Furthermore, the TFN numbers allow for a distinction between lowest, most promising (most frequent) and highest values, which can be linked to patients requiring low attention in terms of consultation time, as well as average- and high-demand patients. To conduct cost calculations for complex patients, the highest possible values will provide an initial estimate. Financial departments of hospitals can benefit from these insights. Moreover, the results of the current study can be used to examine a fair bundle payment rate for an episode of care. A third strength of the study is the broad range of costs (e.g. resources, activities and pharmaceutical) that are incorporated in the analysis, detailing the different factors associated with RA patient care. Furthermore, since the (standardized) patient journey is used as the basis for the calculation, activities performed by staff members not specifically taken into account in the DTC (i.e. ABC) methodology are also included.

A limitation of this study concerns the fact that the care cycle predominantly considered RA costs that are attributable to the hospital. However, the impact of chronic diseases on healthcare expenditures also reaches beyond the hospital silo and involves general practitioners and homecare. A second limitation is caused by the cost pricing system utilized in The Netherlands to reimburse healthcare expenses, as DTC codes do not include all healthcare activities and providers in relation to the healthcare provided, therefore the results of the ABC and (FL)-TDABC models are not fully comparable. This implies that the difference between the ABC and FL-TDABC annual costs could actually be higher than €186 per RA patient. In addition, as previously mentioned, the assumptions underlying the calculation of the DTC prices are not disclosed. Another limitation is related to the FL time estimates of the first consultation and the diagnosis consultation with the rheumatologists. Since it was unclear whether the time estimates included administration and multidisciplinary consultations, the cost differences between the ABC, TDABC and FL-TDABC estimates might be closer in proximity.

In comparison with the existing literature on this subject, a less time-consuming approach is suggested through the use of FL estimates. Staff members were interviewed, and, based on the pooled results, estimates are directly applied in the calculation, without expert panel groups or using the Delphi method to reach consensus [8]. The Delphi method, required to conduct traditional TDABC, involves other limitations concerning generalizability, validation and panelist satisfaction [23]. Additionally, the FL estimates were limited to the more complex activities within the care cycle, where practice variation between patients is most prominent as opposed to analyzing every activity such as a laboratory test.

Applying FL-TDABC has increased the precision and transparency in the costs of the RA care cycle and has provided insight into the cost drivers of the care delivery. Moreover, the model gave insight into the emphasis of the patient journey concerning the expenditures. Performing the (FL)-TDABC steps allows for an in-depth valuation of the practice variation regarding the activities and resources in care delivery, and therefore the model can be recommended as a cost accounting model in healthcare [8].

Potential reductions and optimization in terms of costs might be accomplished by shifting follow-up consultations from the relative costly rheumatologist to, for example, physician assistants and nurse practitioners. A previous study showed that patients are receptive to care delivery by physician assistants and nurse practitioners [24]. This applies, in particular, for the management and monitoring phase, since Dutch patients indicated that the lower-level complex care could potentially be provided by the physician's assistant [25].

With respect to the pharmacy expenses and in the light of minimization of excess healthcare delivery (i.e., waste reduction), close and remote monitoring of therapeutic drug levels may offer a personalized approach and address over- or undertreatment of RA patients [26]. As a result, efficiency gains affecting health outcomes and costs could be attained. Moreover, these results provide management and decision makers with information about the value in terms of cost efficiency related to the healthcare services patients require.

5 Conclusion

Defining healthcare costs through FL-TDABC is relevant from a VBHC perspective, since the time spent, and costs related to the practice variation observed, can be more quickly and adequately linked to the activities, including all relevant providers in the care cycle. Future research will focus on the long-term effects of RA expenditures, differentiating between single morbid and multimorbid patients. In addition, research in terms of a hybrid model considering a more complete allocation of costs to the medical staff within the current system will be investigated.