The randomised controlled trial (RCT) and study setting
Details of the randomised controlled trial have been published (Miranda Filho et al. 2004). Briefly, it comprised 120 patients, aged 12 years or older, recruited from an intensive care unit (ICU) in Oswaldo Cruz University Hospital, in the city of Recife, state of Pernambuco, north-east Brazil, between July 1997 and July 2001. This is the referral hospital for patients with tetanus (accidental or neonatal) and admits virtually all cases of tetanus in the state. Cases of tetanus were allocated to two groups: a control group (62) and a study group (58); placebo was not used, as it would be unethical. Patients allocated to the control group (baseline intervention) received 3000 international units (IU) of immunoglobulin, with preservative, intramuscularly. The study group (intervention group) received the same quantity of immunoglobulin also intramuscularly plus an intrathecal dose of 1000 IU of a human immunoglobulin free of preservatives to prevent irritation of the meninges and avoid the need for corticosteroids. Thus, the difference between the two groups was the exclusive use of the intrathecal immunoglobulin.
Different health outcomes were assessed for different periods of time. Tetanus severity was assessed for a period of 10 days, in both arms of the trial. The decision to limit the follow-up and classification of patients according to their tetanus severity to a 10-day period was for convenience and took into consideration the heavy workload faced by medical doctors working at the referral hospital. The 10-day period, however, was enough to capture fluctuations in the clinical severity of tetanus per patient in the control and study groups, showing clearly the health benefits of both interventions.
Tetanus severity was classified as one of four levels: Grade I (trismus + dysphasia + generalised rigidity, present in more than one segment of the body [head, trunk, arms or legs], - with no spasms); Grade II (mild and occasional spasms, generally after stimulus), Grade III (severe and recurrent spasms, usually triggered by minor stimulus or imperceptible stimuli) and Grade IV (the same features as Grade III + sympathetic nervous system hyperactivity syndrome) (Miranda-Filho et al. 2006). Patients were assessed and classified according to their tetanus severity level at day 0, day 2, day 4, day 6, day 8 and day 10. The assessment was conducted by medical experts working at the hospital and involved in the study. To minimise observation bias, we used a standardised form to collect information that would allow classification according to the levels of tetanus severity; rotated among the doctors involved in the clinical classification of patients; recorded the clinical classification on separate forms which contained no information on the patient's treatment or on previous classifications performed by other doctors; and held periodic meetings with the team to discuss doubts. To compare intrathecal and intramuscular therapies by tetanus severity level as a measure of health benefit, we estimated the number of patients that had clinical progression (improvement or deterioration) during the 10-day period. Clinical progression was characterised as a change in tetanus severity in a specific period of time (0–10 days). It was also expected that fluctuations would occur during the 10-day period regarding tetanus severity: a patient could have been classified as Grade IV on day 2, then as Grade II on day 4, Grade III on day 6, Grade II on day 8 and finally Grade I on day 10. These fluctuations were expected as a characteristic of the disease and were captured every two days by clinical assessment.
Other assessed health outcomes for both arms were hospital stay, respiratory assistance and respiratory infection. Rather than using the 10-day evaluation period, we took into account the entire period of the patient's hospitalisation, in number of days, until their discharge (including those who died during the period of evaluation) to define these outcomes (Miranda Filho et al. 2004). Table 1 summarises the health outcome measures in both groups.
Table 1. Health outcome measures in the two intervention groups
|Health outcomes||Control group||Study group||P-value|
|Clinical progression||N = 60||N = 58|| |
|Improvement||23 (38%)||36 (62%)|| |
|Deterioration||37 (62%)||22 (38%)||0.005|
|Hospital stay in number of days (average)||N = 52||N = 54|| |
|≤15 (8.5)||14 (27%)||23 (43%)|| |
|16–30 (23)||17 (33%)||19 (35%)|| |
|>30 (55.5)||21 (40%)||12 (22%)||0.03|
|Respiratory assistance in number of days (average)||N = 30||N = 20|| |
|≤10 (5.5)||4 (13%)||9 (45%)|| |
|11–20 (15.5)||12 (40%)||7 (35%)|| |
|>20 (41)||14 947%)||4 (20%)||0.01|
|Respiratory infection||N = 62||N = 58|| |
|Yes||42 (68%)||29 (50%)|| |
|No||20 (32%)||29 (50%)||0.07|
Model design, data collection and costing analysis
Costs were primarily calculated as per level of severity of tetanus, as defined in the health outcomes section above. The incremental cost was estimated by comparing the costs by tetanus severity level when patients received the immunoglobulin intramuscularly or intrathecally. Cost items included in the two scenarios are summarised in Table 2. Costs for hospitalisation, respiratory assistance and respiratory infection were calculated as the average cost per tetanus severity level, for the entire period of hospitalisation. All listed costs are likely to be affected by the intrathecal intervention, including the costs for overheads and personnel.
Table 2. Cost items included in the two treatment scenarios
|Control group||Study group|
|Antibiotics to treat respiratory infections||Antibiotics to treat respiratory infections|
|Antibiotics to treat urinary infection||Antibiotics to treat urinary infection|
|Other drugs||Other drugs|
|Gases (respiratory assistance)||Gases (respiratory assistance)|
|Mechanical ventilation (depreciation)||Mechanical ventilation (depreciation)|
|3000 IU immunoglobulin||3000 IU immunoglobulin|
|1000 IU immunoglobulin|
|Training for intrathecal administration of immunoglobulin|
The perspective of the analysis was that of the public health sector. All recurrent and capital costs were calculated. The use of resources for both interventions was based on the results of the RCT published in 2004 and reflected the patient's use of hospital resources (in frequency and quantity) as described in Table 2 (Miranda Filho et al. 2004). Costs, however, were estimated in 2010 local currency prices (Brazilian Real) and then converted to USD values, with an average exchange rate to the USD of 0.5663 Brazilian Real (www.oanda.com).
All cost data were collected from the hospital administrative records. Interviews with medical and non-medical professionals were conducted to determine the frequency and use of resources, when information was not available from the RCT.
Overhead costs were estimated taking into account the value of all contracts of service supplied to the hospital including waste collection, cleaning, building and equipment maintenance, water, telephone and electricity and administrative costs, in a year. As a proxy, we allocated 40% of these contract costs to outpatient activities and 60% for inpatient activities, based on hospital production and the opinion of the hospital's administrative staff. We then divided the cost allocated to inpatients by the total annual hospital patient-days, including hospitalisations due to tetanus. We recognise this is a rough approach, as outpatient and inpatient treatments have different levels of complexity and cost allocation should reflect these differences. However, as this exercise of weighting productions by their level of complexity would require close monitoring of hospital activities for a considerable amount of time, we opted for the simple approach of allocating the total cost by outpatient and inpatient activities. We also added the cost of meals served to tetanus patients during their hospital stay to the cost of overheads. This cost was estimated as the total cost of meals, in a year, divided by total annual hospital patient-days (Drummond et al. 2005). The overhead cost parameter was tested in the sensitivity analysis.
To estimate the cost of personnel per day, we took the amount paid in salaries per year to doctors and nurses working in the referral ICU and then multiplied this amount by the proportion of tetanus patient-days admitted to the ICU (total tetanus patient-days at the ICU divided by the total annual patient- days at the ICU) (Drummond et al. 2005).
The hospital kept records for the annual cost of gases, but did not keep information about the cost of gases by any level of use, for example, per patient or hospital unit. Thus, it was difficult to allocate the gases used for respiratory assistance to patients in treatment. As an alternative, we used the average market price of gases per hour, delivered by different types of private company to hospitals and clinics, and multiplied this by 24 to estimate the average cost of gases per day. This information was supplied by private companies. In Brazil, the price of gases for respiratory assistance in the private and public sector is similar. To calculate the cost of gases, we added the cost per day of a straight-line depreciation for the equipment used for mechanical ventilation, taking into account the life expectancy of the equipment.
The use of antibiotics for the treatment of respiratory and urinary infections, use of other drugs for clinical treatment, consumables and tests were estimated on the basis of the expert opinion of the first author when this information was missing from patients' records. The average cost of drugs per type of treatment was estimated as the cost of drug (in millilitres, milligrams, etc.) multiplied by the dosage, per level of tetanus severity, during the period of hospitalisation. The cost per day was estimated as the total cost of drugs for a specific treatment divided by the average number of days of drugs usage. A similar approach was used to estimate the cost of consumables and tests.
The average cost of an application of 3000 IU immunoglobulin was estimated as the cost of the immunoglobulin plus the cost of one syringe and needle needed for the intramuscular injection of 3000 IU immunoglobulin. The cost of 1000 IU immunoglobulin was estimated as the cost of the immunoglobulin plus the cost of two syringes, a spinal needle 23G appropriate for the application of 1000 IU immunoglobulin, gloves and gauze. A 5% wastage rate for the immunoglobulin, syringes and needles was assumed in the cost calculation (Griffiths et al. 2011). The estimates related to 3000 IU immunoglobulin were made for the control and study groups while those related to 1000 IU immunoglobulin were made just for the study group.
Training was required only for the intrathecal intervention with 1000 IU immunoglobulin, as the 3000 IU is delivered on a routine basis, and no additional training was required. Medical doctors accredited by the Brazilian Unified Health System (SUS) performed the training. We took into account the gross salary paid to these professionals (including productivity bonuses) and divided this amount by the average number of hours they worked in a year. The resulting amount was multiplied by the average number of hours the doctors spent on preparation and delivery of training sessions at the hospital in a year. That amount was divided by the number of tetanus patients in a year, to derive the average cost of training per patient. A total of six training sessions were held, each one lasting, on average, for 60 min, and the average time for the immunoglobulin application was 15 min.
To estimate the total and incremental cost per hospital stay, respiratory assistance and respiratory infection covering the entire period of hospitalisation (shown in Table 7), we adopted the following strategy: hospital stay was calculated as average cost for tetanus for all grades as in Table 5 multiplied by the number of patients multiplied by the average number of days of hospital stay, all summed to the average cost of 3000 UI immunoglobulin for intramuscular use per patient and cost with depreciation of gas equipment per patient; respiratory assistance was calculated as average cost for respiratory assistance for all grades as in Table 4 multiplied by the number of patients multiplied by the average number of days of use of respiratory assistance, all summed to the average cost of 3000 UI immunoglobulin for intramuscular use per patient and cost with depreciation of gas equipment per patient; respiratory infection was calculated as average cost for respiratory assistance plus the average cost to treat respiratory infection for all grades as in Table 4 multiplied by the number of patients multiplied by the average number of days for the treatment of respiratory infection, all summed to the average cost of 3000 UI immunoglobulin for intramuscular use per patient and cost with depreciation of gas equipment per patient. For the study group, for the calculations above described, we added the cost of 1000 UI immunoglobulin for intrathecal use per patient and costs with training per patient.
To test the robustness of the estimates, we used a one-way sensitivity analysis (one parameter is changed at a time), which indicates how the estimates would react to percentage changes in the value of the cost parameters of the model. Using our estimate as a baseline, we varied our cost estimates by plus/minus 10, 20 and 50%.