Table 2.
Summary of base-case model inputs
| Variable | Estimates (95% confidence interval) | Distribution | Source |
|---|---|---|---|
| Probabilities of error | |||
| Error type probabilities with current guidelines (node 1, Fig. 1) | See Table C of Online Supplemental Material |
Dirichlet (no error then error types 1 to 11): Dirichlet (26.3, 0.3, 0.3, 10.8, 0.3, 6.3, 27.3, 0.3, 3.3, 33.3, 0.3, 27.8) |
Previous clinical study [12]. n = 133 |
| Error type probabilities with user-tested guidelines (node 1, Fig. 1) | See Table C of Online Supplemental Material |
Dirichlet (no error then error types 1 to 11): Dirichlet (67.3, 0.3, 0.3, 4.3, 1.3, 0.3, 48.3, 0.3, 6.3, 11.3, 0.3, 3.3) |
Previous clinical study [12]. n = 140 |
| Probability an error is undetected prior to administration (node 2, Fig. 1) |
0.775 (0.718–0.837) 95% prediction interval: 0.655–0.918 |
Log-normal (-0.25, 0.086)a | Random effects meta-analysis of all three randomised controlled trials [24–26] identified by a recent systematic review of the effectiveness of nurse double-checking of medicines administration [27]. The input distribution was based on the 95% prediction interval. See Online Supplemental Material page 13. n = 89,006 |
| Probability an undetected error causes no harm (node 3, Fig. 1) | 0.904 (0.867–0.942) | Beta (217, 23) | Proportion of observed errors from the ECLIPSE study rated as causing no harm (NCCMERP severity rating C) [8]. n = 240 |
| Probabilities of minor, moderate or severe harm for a harmful, undetected error (node 4, Fig. 1) | See Table D of Online Supplemental Material | See Table 3 | Previous clinical study [12] |
| Number of doses of intravenous voriconazole administered using IMG per annum | 4000 | Deterministic | Voriconazole administration and supply data provided by NHS organisations within our perspective - see Online Supplemental Material page 20 |
| Medicine administration time | |||
| Medicine administration time (current guidelines), min | 13.3 (12.5–14.2) | Normal (13.3, 0.41)a | Previous clinical study [12]. n = 133. Converted to a cost by multiplying by the hourly cost of a band 5 nurse (see below) |
| Medicine administration time (user-tested guidelines), min | 11.9 (11.3–12.5) | Normal (11.9, 0.31)a | Previous clinical study [12]. n = 140. Converted to a cost by multiplying by the hourly cost of a band 5 nurse (see below) |
| Medication error costs and QALY decrements | |||
| Cost of an error detected before administration | £0.27 (0.03–2.39) | Log-normal (−1.29, 1.10)b | Karnon et al. [21] (see Online Supplemental Material pages 21–22) |
| Treatment cost for a minor pADE | £124 (94–163) | Log-normal (4.82, 0.14)b | |
| Treatment cost for a moderate pADE | £1252 (1092–1436) | Log-normal (7.13, 0.07)b | |
| Treatment cost for a severe pADE | £1846 (1643–2074) | Log-normal (7.52, 0.06)b | |
| QALY decrement following a minor pADE | 0.004 (0.002–0.006) | Normal (0.004, 0.0011)b | |
| QALY decrement following a moderate pADE | 0.035 (0.017–0.052) | Normal (0.035, 0.0089)b | |
| QALY decrement following a severe pADE | 3.50 (1.87–5.13) | Normal (3.50, 0.83)b | |
| User-testing resources | |||
| Length of pilot user-testing interview, minc | 24 (19–28) | Normal (24, 2.2)a | User-testing study [11]. Converted to a cost by multiplying by sum of the hourly costs (see below) of a band 8a pharmacist (user-tester) and a band 5 nurse (participant) |
| Length of round 1 user-testing interview, minc | 29 (24–33) | Normal (29, 2.5)a | |
| Length of round 2 user-testing interview, minc | 23 (18–27) | Normal (23, 2.1)a | |
| Length of round 3 user-testing interview, minc | 20 (17–22) | Normal (20, 1.2)a | |
| Number of pilot interviews | 4 | Deterministic | User-testing study [11] |
| Number of interviews in rounds 1–3 | 10 | Deterministic | |
| Interview transcription costs, £/spoken min | 1.75 | Deterministic | |
| Length of pilot interview analysis, minc | 94 (82–106) | Normal (94, 6.1)a | User-testing study [11]. Converted to a cost by multiplying by hourly cost of a band 8a pharmacist (see below) |
| Length of round 1 interview analysis, minc | 76 (69–82) | Normal (76, 3.5)a | |
| Length of round 2 interview analysis, minc | 53 (39–66) | Normal (53, 6.9)a | |
| Length of round 3 interview analysis, minc | 50 (46–54) | Normal (50, 2.2)a | |
| Time to revise guides after round 1, minc,d | 265 (226–304) | Normal (265, 20)a | User-testing study [11]. Converted to a cost by multiplying by hourly cost of a band 8a pharmacist (see below) |
| Time to revise guides after round 2, minc,d | 140 (118–162) | Normal (140, 11)a | |
| Time to revise guides after round 3, minc,d | 140 (118–162) | Normal (140, 11)a | |
| User-tester training costs (course cost, trainee time and travel), £e | 562 | Deterministic | User-testing study [11] |
| Equipment costs, £e | 4 | Deterministic | User-testing study [11] |
| Staff costs | |||
| Hourly NHS cost of a hospital pharmacist to carry out user-testing, £ | 65 | Deterministic | Personal and Social Services Research Unit data [28] |
| Hourly NHS cost of a hospital nurse as user-testing participant or to administer voriconazole, £ | 37 | Deterministic | Personal and Social Services Research Unit data [28] |
| Discount rate | |||
| Discount rate | − 0.035 | Deterministic | NICE methods of technology appraisal [23]. |
IMG Injectable Medicines Guide, NCCMERP National Coordinating Centre of Medication Error Reporting and Prevention, NHS National Health Service, NICE National Institute for Health and Care Excellence, QALY quality-adjusted life year, QoL quality of life
aDistribution uses standard error
bDistribution uses standard deviation
cPrevious user-testing study developed IMG guidelines for two medicines, so times halved
dAssumed relative standard error of 8%, (the mean relative standard error for interview and analysis time)
eCost assumed to be shared between ten user tests