In the United Kingdom, the provision of oxygen at home can be prescribed either in cylinders (capacity of 1360 litres) or by concentrator (oxygen extracted from air). Oxygen is prescribed either for treating symptoms or as long term oxygen therapy, when it is used to prolong survival in patients with hypoxaemic respiratory failure.^1,2 For patients prescribed oxygen, it is more practical and cost effective to provide a concentrator if a large number of cylinders are used per month. The Drug Tariff prescribing guidelines and the Monthly Index of Medical Specialities advocate a change from cylinders to concentrator when monthly usage exceeds 21 cylinders,³ a figure reiterated in a recent review of provision of oxygen at home.⁴

We used economic appraisal to determine the level of cylinder use at which it becomes more cost effective to provide oxygen by concentrator at home in Northern Ireland, and to examine the potential cost savings if cylinder use above this level had been replaced by concentrator in 1996.

Economic evaluation is concerned with comparing the consequences of healthcare interventions with their costs. The four main types of economic evaluation all deal with costs but differ in the way the consequences of an intervention are measured and valued.⁵ As there was no reason to presume that medical effectiveness was altered between the two modes of provision of oxygen, we chose cost minimisation analysis. This form of cost effectiveness analysis is used when the consequences of the alternatives are deemed to be equivalent.⁵ In this context, we defined the more cost effective intervention by cost alone.

The provision of concentrators is funded regionally and not from the prescribing budgets of fundholders or regional health boards. Concentrators are installed and serviced by a single contract holder. Each concentrator has a meter to measure use, and the running costs can be reclaimed by the user. The back-up cylinder is provided by the contract holder at an agreed rate as part of the contract.

Table 1 shows the individual component costs for provision of oxygen by concentrator and cylinder. The ingredient cost of a single delivered oxygen cylinder represents 15% of the total cost. Both modes of provision are exempt from value added tax.

Table 1 Individual fixed and variable costs in provision of oxygen by concentrator and cylinder at current costs

As fixed costs do not vary, the level of oxygen usage at which it became cost effective to change delivery mode was dependent on variable costs. In effect, the duration of the oxygen prescription and oxygen usage per day (both flow rate and duration of use) were the key determinants. We performed costings for two hypothetical but not atypical clinical scenarios to illustrate the costing methodology and to show how prescription duration and daily usage affected total cost. We undertook threshold analysis to indicate the values of key variables that would justify revision of prescribing guidelines. For all costings we assumed there were no concentrator breakdowns (an unusual event) and that all deliveries, comprising three cylinders per delivery, were within 6 miles.

Individual concentrator prescriptions are issued from a regional department but cylinder prescriptions are not collated centrally. The only data available were the total ingredient cost and total number of prescriptions in Northern Ireland. The cost of this service varies substantially depending on individual prescription, frequency of deliveries, and distance delivered, and thus it was impossible to determine the actual cost of provision of oxygen. We used sensitivity analysis to provide a range of estimates on the basis of the assumptions about the values that particular variables were likely to take, and we used this to give an estimate of the potential savings.

Table 2 shows the costs associated with this scenario. At this usage a concentrator is a more cost effective option than cylinders. As usage falls the total costs for each mode of provision approximate, but it would require a reduction in cylinder usage to three for the month (flow rate 2 l/min) before the concentrator became more expensive £66.26 ($106.02) v £59.62 ($95.39)). Using four cylinders for the month at 2 l/min flow rate, the cylinder would cost £84.58 ($135.33) whereas the concentrator would cost £67.05 ($107.28). Thus three cylinders used for 1 month at a flow rate of 2 l/min represent the cost effective cut off point beyond which the provision of oxygen will always be cheaper by concentrator.

Table 2 Fixed and variable costs for 5 hours’ oxygen usage per day for 1 month at a flow rate of 2l/min for both concentrator and cylinder

Table 3 shows the costs associated with this scenario. With this amount of use a concentrator is more cost effective than cylinders. Again, as usage falls the total costs for each mode of provision approximate, but it would require a reduction in oxygen usage to four cylinders every 3 months over the 2 years before the concentrator became more expensive than cylinder provision (£573.53 ($917.65) v £486.23 ($777.97)). The cost of provision of five cylinders every 3 months is £596.01 ($953.62) and similar concentrator usage would cost £579.87 ($927.79). Thus, with oxygen usage over a longer period, cylinder usage becomes more expensive and the cost effective cut off point relates to a smaller number of cylinders used per month, indicating the importance of duration of prescription.

Table 3 Fixed and variable costs for 90 minutes’ oxygen usage per day for 2 years at a flow rate of 2l/min for both concentrator and cylinder

If the duration of prescription is unknown (which is generally the case at prescription issue) it is still possible to determine a cost effective cut off point on the basis of oxygen usage per month. The cut off point for provision of two cylinders per month at a flow rate of 2 l/min is 6 months (concentrator £186.81 ($298.90) (including 6 month service as provision is ongoing) v cylinder £186.03 ($297.65)) since again beyond this it is always cheaper to provide equivalent oxygen usage by concentrator.

This cut off point for total cylinder usage is also dependent on flow rate, as at lower flow rates and similar cylinder usage per month the running costs of the concentrator for the same provision of oxygen rises. Thus, given the usage of two cylinders per month at a flow rate of 1 l/min the cost effective cut off point is 12 months (concentrator £344.63 ($551.41) v cylinder £353.29 ($565.26)). In annual terms, a concentrator is always cheaper if more than 22 cylinders are being consumed per year (annual cost of 22 cylinders or concentrator equivalent: concentrator £341.46 ($546.34) v cylinder £339.33 ($542.93)).

We conclude that if more than three cylinders per month are being used, independent of flow rate or duration of prescription, it is always cheaper to prescribe a concentrator. If the duration of prescription is likely to be 12 months or longer it is always cheaper to prescribe a concentrator when two or more cylinders are being used per month whatever the flow rate.

Table 4 Number of patients, total number of oxygen cylinders prescribed in Northern Ireland for 1996, and total ingredient cost at current price

Table 5 Costing of cylinder provision in 1996 (assuming 23 cylinders per patient using current costs) and two alternative provision options on basis of actual cylinder usage and number of prescriptions in 1996. Costs are in £

Our study shows that in Northern Ireland it is currently cost effective to provide oxygen by concentrator when the patient uses three or more cylinders per month, independent of the duration of the prescription. If the period of oxygen usage exceeds 12 months, it is more cost effective to provide oxygen by concentrator when the patient uses two cylinders per month. The current Drug Tariff prescribing guidelines, advocating that provision of oxygen by concentrator becomes cheaper when 21 cylinders are being used per month, are therefore inaccurate in Northern Ireland.

We examined the range of potential savings in Northern Ireland. We showed that current information on cylinder prescription was insufficient as individual oxygen prescriptions, frequency of oxygen usage, and delivery costs for oxygen rather than a crude average and ingredient cost, were crucial in determining the savings (or costs) that could be achieved by more widespread prescription of concentrators. There should be little resistance from general practitioners to prescribing more concentrators as this service is funded centrally, and fundholding practices may make substantial savings. Pharmacists may, however, be disappointed as a move to more concentrators is likely to cause some loss of income.

Concentrators are unobtrusive, reliable, and convenient and can be used in multiple outlets in the home. Patients generally prefer this mode of provision of oxygen, although for some the absence of regular deliveries by the pharmacist may increase feelings of isolation. The alternatives we examined were those available for provision of oxygen in the home in the United Kingdom on NHS prescription. Liquid oxygen is not available on the NHS, and although it is a more flexible source of oxygen at home owing to its portability it can only be obtained if purchased by the patient. The costs of contracts for provision of concentrators are similar throughout the United Kingdom and are equivalent to other European countries.

The major assumptions in our cost analysis were that pharmacists transported three cylinders per delivery and that all deliveries were within 6 miles. The delivery of three cylinders is advocated in the Drug Tariff prescribing guidelines: “GPs are asked to cooperate in an effort to make this part [supply] of the oxygen service more cost effective. If more than one or 2 (cylinders) are regularly required for a particular period, prescribing in multiples of 3 cylinders would reduce savings by reducing the number of journeys.” Delivery of three cylinders by pharmacists would seem to be standard practice (personal communication). A pharmacist could deliver more cylinders, which would affect the cost of the service. However, even if six cylinders were provided with each delivery in scenario 2, it would still be a more expensive option. It is also possible that the delivery service could cost more if one cylinder were delivered on each occasion. The assumption that all cylinders are delivered within a 6 mile radius is almost certainly inaccurate, particularly in rural parts of Northern Ireland. The effect of this assumption was to reduce the cost of provision of cylinders making our conclusions more robust.

The cost of concentrator usage was based on the contract of the current provider. More widespread use of concentrators would have an effect on contract price, which would probably fall, altering the cost effective cut off point to fewer cylinders used per month.