The term "INR controls" is below defined as all aspect involved in the monitoring of warfarin, including preparations and follow-up.
Frequency of INR controls
The frequency of INR controls during the initiation phase and during established treatment was derived from two retrospective studies of electronic patient records [3, 13, 14]. Both studies were performed in Stockholm, where patients on warfarin routinely are managed in primary care.
One study was performed on patients with CAF during the initiation phase of warfarin treatment (the first three months of treatment) [13]. Twelve primary health care (PHC) centres from five different health care districts in Stockholm County with a registered population of 203, 407 individuals included 144 patients.
In the other study, five PHC centres with a registered population of 75,146 participated. Twenty-five patients with CAF, who had received a minimum of 30 days treatment were randomly selected for a detailed review of the clinical management of warfarin treatment [3, 14]. Of those, five patients had been treated for less than three months, one had been on treatment for 11 months and 19 had been on treatment for more than one year. All patients in the initiation phase of treatment were excluded and data from the remaining patients (n = 20) were used to represent the frequency of INR controls in established patients, defined as treatment for 12 months or more. Information on the frequency of INR controls between four and 11 months treatment was not available and an assumption of a linear decline was applied, based on available data on the frequency of INR controls during month three and month 12.
Resources consumed per INR control
The type and number of resources consumed at INR controls were investigated in two Delphi-panel studies [15]). In such panels standardised techniques are used to systematically collect and collate informed judgments from a group of experts on specific questions or issues [16–20]. The respondents are anonymous to each other and the study is performed in different rounds.
The respondents in both Delphi-panels were health care personnel. One was performed in three rounds and had the objective to study the resource use of patients whose INR controls were carried out onsite at a PHC centre. Thirty-four general practitioners (GPs) and 10 registered nurses (RNs) from 34 PHC centres in Stockholm participated. The other Delphi panel investigated the resource consumption for patients managed in home care. It was performed in two rounds and 49 district nurses (DNs) from all over Sweden were enrolled.
In the application of the Delphi technique in the studies referred to in this article, the expert panel members were randomly selected from GPs and RNs in primary care in Stockholm and DNs in Sweden respectively. The panel members were asked to estimate patient-related time defined as time spent in preparation before the INR control, direct time spent together with the patient and follow-up in connection with the control. Total patient-related time spent by GPs and RNs respectively was regarded as a reflection of resources used in INR controls, apart from the work undertaken by laboratory staff and the cost of transportation to and from patients in home care, which were calculated separately. The costs were estimated by multiplying the various resources by their respective unit cost.
Unit cost calculation
Unit costs were based on data from 2003, estimated from the perspective of the health care provider and expressed in SEK (SEK 1 = € 0.11). Where costs have been reported in $ or ₤ in publications referenced in this article, the following exchange rates have been used; $1 = SEK 7.61 and ₤1 = SEK13.88. All exchange rates are as of June 30, 2004.
Cost of a patient-related hour
Gross payroll expenses were defined as the average annual costs (salary, security fees and pensions) of GPs, RNs and DNs in the County Council of Stockholm in Sweden. It was SEK 868, 977 for a GP, SEK 382,508 for an RN and SEK 412, 482 for a DN.
The average annual working hours in 2003 for full-time employees in the health care sector was 1,576, which was derived from a labour force survey performed by Statistics Sweden (unpublished data). The patient-related time was estimated to be 73% for the GPs. This figure was based on a survey performed by The Swedish Institute for Health Economics in 1999–2000 in which four different specialities were represented including GPs (n = 81) [21]. GPs reported they spend 50% of their total working-hours in face-to-face consultations with patients, and another 23% of their time for administrative work directly related to patients. This means 73% of the working-hours are allocated to patient-related work. We estimated the patient-related time for RNs and DNs to 85% based on the assumption that patient-related time is greater for nurses than for GPs.
Hence the average annual working hours of patient-related time (Ann_pat_TIME) was estimated at 1,150 (0.73*1576) for GPs and 1,340 (0.85*1576) for RNs and DNs.
The gross payroll expenses of a patient-related hour (Gross_payroll_exp_pat_hour) was
Hence the "Gross_payroll_exp_pat_hour" was SEK 755 (868,977/1,150) for GPs, SEK 286 (382,508/1,340) for RNs and SEK 308 (412,482/1,340) for DNs.
The balance-sheets of PHC centres include gross payroll costs as well as overhead and administrative costs such as rental and maintenance of the localities, capital costs and telephone. These costs were denoted "other costs" and were added to the payroll expenses to obtain a total cost of a patient-related hour. The accounts of three PHC centres, one big (Gustavsberg), one middle-sized (Österåker) and one small centre (Nyby), showed the average of "other costs" was 40% of total cost.
The following formula was applied for calculation of the total cost of a patient-related hour including "Gross_payroll_exp_pat_hour " as well as "Other costs":
Thus the cost of a patient-related hour based on "Other costs" being 40% was SEK 1259 (755/0.40) for a GP ('GP_hour_COST'), SEK 476 (286/0.40) for an RN ('RN_hour_COST') and SEK 513 (308/0.40) for a DN (DN_hour_COST).
Sample and transportation costs
The work undertaken by laboratory staff was not investigated in the Delphi panels. Therefore, tariff prices for taking the sample (SEK 80) and the analysis of the sample (SEK 42) were used in the calculations. These costs were derived from Nyby PHC centre.
The cost for car transportation was estimated to be SEK 2.50 per kilometre for a small or mid-sized car (Swedish Consumer Agency in Stockholm).
Estimate of costs
Cost of INR controls at primary care centres
PHC centres were divided into such centres where GPs were routinely assisted by an RN in the management of INR controls for warfarin patients and such centres where GPs did not co-operate with an RN on a regular basis for these controls. All costs are expressed as mean costs.
The cost of an INR control for GPs with an RN ('GPRN_control_ COST') was
['GP_hour_COST' * ('GPRN_TIME' + 'GPRN_exTIME') +'RN_hour_COST' * ('RN_TIME' +'RN_exTIME')] + 'Sample_COST' + 'Analysis_COST'
where
'GP_hour_COST' and 'RN_hour_COST' have been defined above,
'GPRN_TIME' = Mean time GPs, routinely assisted by a nurse, spend on one INR control,
'GPRN_exTIME' = Mean extra time for GPs, routinely assisted by an RN, when a patient does not appear for a scheduled visit.
'RN_TIME' = Mean time nurses spend on one INR control,
'RN_exTIME' = Mean time for an RN when a patient does not appear for a scheduled visit.
'Sample_COST' = Cost of taking the blood sample by laboratory staff and
'Analysis_COST' = Cost of analysing the blood sample.
Similarly, the cost of an INR control when the GP was not routinely assisted by an RN (GP_control_COST) was
* ['GP_hour_COST'* ('GP_TIME' + 'GP_exTIME')] + 'Sample_COST' + 'Analysis_COST'
where
'GP_hour_COST', 'Sample_COST' and 'Analysis_COST' have been defined before and
'GP_TIME' = Mean time GPs, not routinely assisted by a nurse, spend on one INR control,
'GP_exTIME' = Mean extra time for GPs, not routinely assisted by an RN, when a patient does not appear for a scheduled INR control.
Accordingly the total cost of an INR control when the patient is managed onsite at a PHC centre ('PHC_control_COST) was
('GPRN_PROP' * 'GPRN_control_COST') + (GP_PROP * GP_control_COST)
where
'GPRN_control_COST' and 'GP_control_COST' have been defined before and
'GPRN_PROP' = Share of PHC centres where doctors are routinely assisted by a nurse,
'GP_PROP = Share of GPs, not routinely assisted by a nurse,
Cost of INR controls in home care
The total cost per INR home care visit ('HOME_control_COST') was:
['DN_hour_COST' * ('DN_TIME'+ 'DN_exTIME')] + (Transp_COST * Transp_KM) + 'Analysis_COST'
where
'DN_hour_COST' and 'Analysis_COST' have been previously defined and where
'DN_TIME' = Mean time district nurses spend on one INR control visit,
'Transp_COST' = Cost per kilometre by car (Swedish Consumer Agency),
'Transp_KM' = Mean distance per INR control, return trip and
'DN_exTIME' = Extra time per INR control due to INR control visits undertaken in vain.
Costs of INR control in primary care
Total cost of an INR control undertaken in primary care ('TOT_control_COST') was
('PHC_PROP' * 'PHC_control_COST') + ('HOME_PROP' * 'HOME_control_COST')
where
'PHC_control_COST' and 'HOME_control_COST' have been previously defined and
'PHC_PROP' = Proportion of INR controls undertaken at a PHC centre and
'HOME_PROP' = Proportion of INR controls undertaken in the patient's home.
Cost per patient
Total cost of INR controls during the initiation phase (first three months of treatment), during the first year (including the initiation phase) and during subsequent years of treatment was obtained by multiplying TOT_control_COST with the monitoring frequencies observed in or derived from the two retrospective studies on electronic patient records mentioned above [4, 13, 14].
Sensitivity analyses
Although this analysis provides the most accurate assessment of the economic impact of INR controls in primary care available to date, a number of uncertainties in the estimates of resource use and unit costs were identified and explored in a series of sensitivity analyses. The following analyses were performed: (i) reduction/increase of the time GPs (GPRN_TIME and GP_TIME and DNs (DN_TIME) used for INR controls, (ii) the proportion of home visits ('HOME_PROP') was lower/higher, (iii) 'Other costs' were lower/higher, (iv) the costs incurred by laboratory staff ('Sample_COST') and analysis of the sample (Analysis_COST') were lower/higher.