Costing methodology

Evaluating how much it costs GP practices to deliver vaccinations has never previously been attempted. Understanding how difference in practice structure and function impact on the financial base of a practice is likely to be key in evaluating differences in performance between different organisational models and will provide an explicit quantitative comparison considering aspects of vaccine programme delivery, but also wider contextual factors at play within the practice. Costing in

85 primary health care services is notoriously challenging due to the complex and variable nature of the type of services delivered. The additional challenge in England is that each GP practice is an

independent organisation and, in essence, a private business, although usually with the NHS providing the vast majority of funding. This means that data on costs and expenditure at practice level are very difficult to come by, as they are neither centrally collected nor publicly available. The specific challenge with vaccination is that, in most cases, activities are both distributed amongst multiple different staff groups, including doctors, nurses, managers and administrators; and have a wide temporal distribution with activity undertaken both unevenly throughout a working week as well as throughout the year. This makes both apportioning capital and operational costs and also staff time to vaccination particularly difficult without relatively extensive data collection in relation to vaccination activity within a practice.

5.3.1 Costing methods

To undertake costing at practice level costs must be identified for capital costs (buildings, land, equipment), overhead costs (utilities, maintenance, waste disposal consumables), and staff costs through salaries (Bowling, 2009b). The costing methods employed then need to generate sufficiently granular data to allocate these costs to vaccination activity.

A complex system of tariff payments is used to remunerate secondary care organisations in England based on activity. This is managed by NHSE and NHS Improvement and is based on costs reported via hospitals and other provider organisations, which are used to develop a national tariff, for which providers are remunerated based on their recorded activity (NHS England, 2016a; NHS Improvement, 2018b, 2018a). However, as GP practices exist as independent organisations contracted to provide services by NHSE, no such equivalent costing exists, and the original payment thresholds in the contracting arrangements have been set through mutual negotiation or rolled over at historical levels generated by past activity. Therefore, to identify costs associated with vaccination, all the information needs to be gathered in a suitable format from each included practice, as costing data are not

otherwise available.

5.3.1.1 Measuring time and activity

To allocate costs, both overheads and staff salaries, time spent by staff on vaccination needs to be known. Since the Industrial Revolution, organisational management researchers have been interested in how workers spend their time. Perhaps surprisingly, the most widely used methods for evaluating time spent on specific activities have not changed significantly in the last one hundred years. There are broadly three methods that have been widely employed in business, industry and, more recently, in healthcare settings. The first is self-reported timekeeping and the second is observational work sampling. Self-report simply requires workers to write down the time spent on specific activities,

86 either prompted or un prompted, whereas in work sampling the recording of time spent on specific activities is undertaken by a researcher through observation at specific intervals during a working period. One observer may then be responsible for measuring activity amongst multiple different staff members, if this activity is taking place in a well-defined work area. A study conducted amongst nurses in an Australian hospital compared these two methods and found differences in recorded outcomes and challenges with each method (Ampt et al., 2007). Nurses found it difficult to fill in the self-reporting forms while busy with clinical tasks, which led to an under-reporting of activities. Staff found the observational method preferable and this also led to a higher volume of activity

measurements, however this requires trained researchers available to observe clinical staff and there is a risk of creating a Hawthorne effect, where activity changes as a result of being observed.

The third is a time-and-motion study, which is a more detailed method of observational work

sampling that usually involves an observer measuring all activity during an entire work period using a stopwatch. In this case the observer will usually follow a single staff member and accurately record all activity. This does produce a highly accurate picture of work activity, however it also has a higher risk of the Hawthorn effect where the researcher may change staff behaviour due to their close observation (Finkler et al., 1993). It also necessitates observing a smaller sample of staff due to the high level of research resources required. Time-and-motion studies have been used extensively in health services research in recent years, particularly in hospital practice. However, a recent review of these studies highlighted significant methodological variability between studies, which made comparisons and aggregation of results difficult (Lopetegui et al., 2014). This method has also been used in primary care. In one study, researches evaluated the effect of the implementation of an electronic care record on doctors’ time in 5 clinics (Pizziferri et al., 2005). This study required 7 research assistants to undertake the observations. In the context of evaluating the delivery of routine vaccinations in GP practices, a time-and-motion study would not be feasible, because, at most practices, vaccination appointments are randomly distributed throughout the week. Therefore, to capture time-and-motion data, a researcher would have to sit through many days of clinics to capture a relatively small number of vaccination appointments. Similarly, the administrative work is not distributed evenly, and would require administrators to plan their vaccination work in advance to enable the researcher to visit at the appropriate time, which may affect the outcome. In practices that do run specific vaccination clinics, most often there are multiple staff working simultaneously, including nurses and administrators, thus, either requiring multiple researchers, or multiple visits to a clinic to record activity from each

different staff member. Overall, although time-and-motion studies may provide highly accurate data on activity on vaccination within GP practices, due to resource constraints it is not possible to use these methods for this study. In addition, while these methods identify time and activity, they do not provide a method for allocating costs to these activities.

87

5.3.1.2 Time-Driven Activity Based Costing

Activity Based Costing (ABC) is a method for allocating costs to complex processes within

organisations to identify cost drivers and opportunities for lowering costs while improving quality. It was developed by Cooper and Kaplan for use within business and industrial contexts as a method that requires relatively low resources and is transparent while remaining accurate (Cooper and Kaplan, 1999). It focusses on identifying all costs for delivery of organisational outputs and allocating these to different processes and activities based on the time taken to complete them by actors within the system. Having been widely applied in the private sector, it was relatively swiftly adopted by public sector organisations in the USA, including health service providers (mainly hospitals). The cost of delivering vaccination at primary care level has previously been evaluated in New Zealand on two occasions using ABC methods and found to be feasible and capable of producing accurate and useful results (McLeod, Bowie and Kljakovic, 1998; Turner et al., 2009). More recently ABC has been updated to reduce the administrative burden of data collection and is now termed Time-Driven Activity Based Costing (TDABC) (Kaplan and Anderson, 2007; Kaplan et al., 2014). In part this development was to make TDABC more suitable to healthcare contexts, specifically to improve the ability for healthcare organisations to undertake costing as part of the new drive towards Value Based Healthcare, with a focus on achieving maximum health outcomes for given inputs (Porter, 2009). As a result, TDABC has since been extensively applied in a wide range of healthcare contexts, including both primary and secondary care, as highlighted in a recent review by Keel et al., (Keel et al., 2017). The review identified that TDABC has been used to improve service delivery, allocate payments to providers, and that primary study authors had chosen this method due to its ability to accurately capture and allocate costs and describe the process of delivering care within complex healthcare organisations. TDABC has mainly been used in secondary care, however there are also a smaller number of studies in primary or community care settings. One successfully evaluated the

implementation of a physical activity pathway in GP practices in England to identify the costs to practices associated with the programme (Boehler et al., 2011).

TDABC follows a seven-step process (Kaplan and Porter, 2011):

• Step 1 is to define the condition under analysis, which should include clear definitions of the start and the end of the process under evaluation.

• Step 2 is to define the Care Delivery Value Chain (CDVC). This is a visual map that details the core activities of the process under evaluation across a single cycle of care.

• Step 3 is to develop a process map. This is more detailed than the CDVC and includes all the paths that a patient may follow through the process to any specific end point.

88 • Step 4 is to generate time estimates for each stage in the process. This can be done through

any method (including those described in the paragraph above) and may be extracted from electronic patient records, self-report by staff, or through observation from researchers. • Step 5 is to estimate the cost of providing a unit of care as per the CDVC and involves

identifying the costs associated with the process and allocating these to the time estimates derived in step 4.

• Step 6 is to estimate the Capacity Cost Rate (CCR) for the resources described in step 5. The CCR is a cost per unit time of the available resources, whether it is being used or whether it remains available for use as part of the process.

• Step 7 then completes the process by calculating the total cost of one cycle of patient care. Therefore, the outputs from the analysis are the CDVC, a process map, description of time per activity and both overall and individual unit costs for stages within the process. TDABC provides a

comprehensive approach to identifying process, activities, time spent, and associated costs of

interventions delivered in complex healthcare settings, including in primary care, with a relatively low burden on both practice staff and researcher resources, particularly as it allows flexibility of

determining how time and activities are recorded. Therefore, I have selected to use this method to identify aspects of implementation and associated costs of routine vaccination at GP practices as part of this study. Applying this method across different practices will also enable a comparison of both differences in implementation and associated costs.