Rapid tests for group A streptococcal infections in people with a sore throat

A decision tree was created to simulate the potential care pathways associated with using rapid tests and clinical scoring tools, compared with using clinical scoring tools only (current practice), in people with acute sore throat.

The economic analysis was from the UK NHS and personal social services perspective. A 1‑year time horizon was used to see the effect of rare but serious complications of strep A infection on costs and outcomes (a shorter time frame of 14 days was used in sensitivity analyses). No discounting was applied to costs and benefits because of the short time horizon. Because of the lack of published data, the models did not consider wider public health benefits such as the potential effect on antimicrobial stewardship or onward transmission rates.

A prevalence of 22.6% was used for adults, based on the study by Little et al. (2014). The study enrolled patients aged 3 years or older in UK primary care. For children, an estimate of 30.2% was assumed, based on the median of 3 non-UK studies of children in primary care.

The accuracy estimates for the Centor clinical scoring tool were taken from the meta-analysis by Aalbers et al. (2011). It focused on Centor to predict strep A pharyngitis in adults (15 years or older) in primary care. At the Centor threshold of 3 or more, the sensitivity was estimated as 49% (95% CI 38% to 60%), and specificity as 82% (95% CI 72% to 88%). There were no studies reporting the accuracy of the FeverPAIN clinical scoring tool so it could not be modelled.

The accuracy estimates for the rapid strep A tests were from the systematic literature review done by the EAG. The sensitivity of the rapid tests ranged from 68% to 100%, and the specificity from 79% to 100% (see table 4 and table 5). The estimates of accuracy based on unpublished manufacturers' data or FDA reports were consistently higher than the estimates from the published peer-reviewed studies. Therefore, the economic models based solely on manufacturers' test accuracy data should be interpreted with caution.

Treatment-related probabilities and complication rates used in the models are in table 6.

The health impact of each pathway was expressed in QALYs. These were calculated by subtracting the disutilities associated with treated and untreated strep A infection, complications of strep A infection and adverse effects of penicillin (see table 7) over 1 year from the mean baseline utilities. The mean baseline utilities in the models were based on a general UK population: 0.863 for adults and 0.94 for people under 25 years (Kind et al. 1998). The latter is the closest age group to children and therefore was used as a baseline utility in the children's models. Mean disutilities were based on published literature (Neuner et al. 2003; reported as quality-adjusted life days), by converting quality-adjusted life days to utility decrements.

Costs were calculated using 2017/18 prices. The total costs for each strategy (current practice and rapid tests) include GP consultations, antimicrobial therapy, and managing strep A infection-related complications and adverse effects of penicillin (see table 8).

Cost data were available for 14 of the 21 rapid tests in this assessment (see table 9). The cost of testing also accounted for:

• Additional GP time needed to process the test, ranging from 5 to 12 minutes depending on the test.

• Apportioned cost of analyser or test cassette reader (that is, cost of analyser or reader adjusted for its average life span and the average number of samples analysed).

• Cost of the microbiological culture of throat swabs (£8 per sample) to confirm negative test results, when needed.

The model was created for adults in primary care and then adapted for children and secondary care:

• In current practice, antibiotic prescribing (immediate, delayed or no prescribing) is based on the Centor score.

• In the rapid test cohort, people with a Centor score of 3 or more are offered the rapid test. Antibiotic prescribing decisions (immediate, delayed or no prescribing) are based on the test results.

• Of people offered delayed prescription, 46% use their prescription.

• There are 1.3% to 1.5% of people with strep A infection who develop complications, depending on whether or not they had antibiotics.

• People who take antibiotics are at risk of penicillin-related adverse effects (2% have penicillin-induced rash and 0.01% have penicillin-induced anaphylaxis).

• When recommended by manufacturers, negative results for rapid strep A tests were followed up with a microbiological culture or throat swabs to confirm the results.

The model for adults in secondary care was adapted from the adult primary care model by excluding the cost of the initial GP consultation. Also, it was assumed that all rapid tests could be done in the standard time allocated for secondary care appointments. The accuracy of rapid tests was assumed to be the same as in primary care (because of the lack of specific data in secondary care) except for 3 tests for which the sensitivity estimates from secondary care were available: OSOM Strep A test (94%), QuikRead Go Strep A test kit (87%) and the Alere TestPack Plus Strep A (90%). All other assumptions and inputs are the same as in the primary care model.

The model for children in primary care was adapted from the corresponding adult model by adjusting the prevalence of strep A infections from 22.6% to 30.2%, and using the accuracy estimates from studies in children whenever these were available (see table 5). The costs of treating peritonsillar abscess and related complications in children were assumed to be lower than in adults (£1,420.50 compared with £1,571.28), based on the NHS reference costs for both age groups.

The test accuracy data for children in secondary care were assumed to be the same as in primary care (because of the lack of specific data in secondary care), except for 3 tests for which the accuracy estimates from secondary care were available: OSOM Strep A test (test strip; sensitivity: 94%, specificity: 97%), QuikRead Go Strep A test kit (sensitivity: 87%, specificity: 78%), and Alere TestPack Plus Strep A (sensitivity: 77%, specificity: 97%).

In the base-case adult primary care model, current practice dominated (that is, current practice was more effective and cheaper than the testing strategy) 2 tests: the Clearview Exact Strep A cassette and dipstick. The ICERs for the remaining 12 tests ranged from £1,353,677 to £6,059,081 per QALY gained, compared with current practice (see table 10). Costs and QALYs were multiplied by 1,000 because of the very small incremental QALYs.

The results of the base-case adult secondary care model were in line with the results of the adult primary care model, but the ICERs were much lower (see table 11).

In both models for children, current practice dominated 4 tests: the Clearview Exact Strep A cassette and dipstick, QuikRead Go Strep A test kit and Alere TestPack Plus Strep A cassette (see table 11). In the children's primary care model, the ICERs for the remaining 10 tests ranged from £1,762,306 to £7,893,857 per QALY gained, compared with current practice. In the children's secondary care model, the ICERs for the remaining 10 tests ranged from £65,122 to £5,723,279 per QALY gained, compared with current practice.

The results of the probabilistic sensitivity analysis mirrored the results of the deterministic base-case analysis in all models.

The probability of a rapid test being cost effective was 0 in all 4 models, regardless of the rapid test used.

A range of scenario analyses was done. For the adult primary care model, none produced ICERs that were around or below £30,000 per QALY gained, compared with current practice.

For the adult secondary care model, changing the rate of penicillin-induced anaphylaxis from 0.01% (Neuner et al. 2003) to 0.64% (Van Howe and Kusnier 2006), resulted in 6 rapid tests dominating current practice (that is, testing was cheaper and more effective than current practice). These were the 5 NADAL tests and Alere TestPack Plus Strep A. The ICERs for 4 tests (2 Clearview Exact Strep A tests and 2 Strep A rapid tests from Biopanda) decreased to around or below £30,000 per QALY gained, compared with current practice.

In addition, for the adult secondary care model, the ICERs for the 5 NADAL tests decreased to around or below £30,000 per QALY gained, compared with current practice, for the following assumptions:

• changing the Centor threshold for starting antibiotics and testing to 2 or more (ICERs: £30,230 to £69,690 per QALY gained)

• changing the Centor threshold for starting antibiotics and testing to 1 or more (ICERs: £22,220 to £56,190 per QALY gained)

• lowering the prevalence of strep A infection to 10% (ICERs: £20,628 to £53,506 per QALY gained)

• doubling the rate of penicillin-related rash to 4% (ICERs: £8,913 to £32,557 per QALY gained)

• doubling the utility decrement of penicillin-induced rash (ICERs: £21,309 to £44,953 per QALY gained).

For the children's primary care model, no scenario analyses produced ICERs that were around or below £30,000 per QALY gained.

The scenario analyses for the children's secondary care model largely mirrored scenario analyses for the adult secondary care model, except that changing Centor threshold for starting antibiotics and testing to a score of 2 or more had no major effect on the ICERs.

In addition, several analyses favoured testing strategies, and all or some of the tests dominated by current practice in base-case analyses were no longer dominated. However, the ICERs were around or above £100,000 per QALY gained:

• doubling the complication rate of treated strep A infection to 2.6%

• halving the complication rate of untreated strep A infection to 0.75% (children's primary care model only)

• halving the utility decrement of untreated strep A infection

• doubling the utility decrement of treated strep A infection

• changing the accuracy estimates to the lower confidence limits for both the rapid test and Centor clinical scoring tool (children's primary care model only).

Several scenario analyses favoured current practice. In all 4 models, doubling the utility decrement associated with untreated strep A infection resulted in an additional 2 to 4 tests being dominated by current practice, compared with base-case results. These tests were the Strep A rapid test cassette and the test strip from Biopanda (all models), the Alere TestPack Plus Strep A cassette (adult primary and secondary care models) and the QuikRead Go Strep A test kit (adult secondary care model). In the adult secondary care model, the following assumptions also resulted in additional tests being dominated by current practice:

• increasing the prevalence rate to 35.9%

• halving the complication rate of treated strep A infection to 0.65%

• doubling the complication rate of untreated strep A infection to 3%

• halving the rate of penicillin-related rash to 1%

• halving the utility decrement of treated strep A infection

• halving the utility decrement of penicillin-induced rash

• doubling the utility decrement of an abscess.