Table 1.
Feasibility outcomes: summary of prior assumptions and actual findings.
| Outcome | Prior assumption | Actual | Comment |
|---|---|---|---|
| Number of participating prisons | 11 prisons in Scotland 14 prisons in England | 16 prisons in England (15 open, 1 closed to recruitment) | Consistent with expectation |
| Target participant accrual | 5600 participants Original sample size requirement was up to 10% of 56 000 participants | Target revised down to 2800 for England due to non‐participation of prisons in Scotland and Wales
February 2014: interim target revised to 1500 by 31 August 2014 Actual accrual at 31 August 2014 was 1392 participants 1500 participants achieved on 8 October 2014 Final accrual at 8 December 2014: 1685 participants |
Early cessation, see Fig. 3 |
| Consent for randomization by eligible prisoners | 75% | Based on screening logs, the consent‐rate for randomization among eligible prisoners was 72% (1283 of 1777); 95% CI = 70–74% | Upper 95% confidence limit is just short of our prior expectation |
| Consent to returned prisoner self‐questionnaire (RPSQ) | Prior expectation of 75% | Consent to complete the RPSQ was given by 85% of participants (1417 of 1676); 95% CI = 83–86% | Above expectation |
| Consent to secondary randomization in the telephone‐contact ancillary study | Prior assumption of 50% | Consent to take part in the telephone contact study was provided by 56% (946 of 1676); 95% CI = 54–59% | Above expectation |
| Number of RPSQs | 333 recidivist self‐questionnaires expected from 2500 randomized and released participants, so we expected 333 of 2500 × 1557 = 207 RPSQs | 218 received from 1557 randomized and released participants | Consistent with expectation |
| Carriage rate in first 4 weeks after release | 75% | RPSQ 71% (80 of 112) 95% CI = 63–79% | Consistent with expectation |
| Someone else present at overdose | 80% | Based on RPSQs: 53 of 205 recidivists (26%; 95% CI = 20–32%) of recidivists reported having injected when alone, and had done so on a mean of 6 of 14 days. Hence, 95% CI for someone else present is 68–80%
Telephone questionnaire: heroin use in the past 3 days was reported by 31 of 81 telephone contacts (38%), 10 of whom had injected when alone (12%; 95% CI = 5–20%). If the past 3‐day rate is taken as representative of the rate throughout the first 4 weeks, then 95% CI for someone else being present at injector's overdose is 80–95% Both consistent with our prior expectation that someone else is present at 80% of opiate overdoses Pooled estimate (based on weights 61% and 39%) is 79%; 95% CI = 75–84% |
Consistent with expectation |
| Telephone questionnaire phone contacts in the first or second fortnight | Based on the (probability of consent) × (probability of being randomized to telephone contact given consent) × (probability of contact given randomized to telephone contact) = 1/8 × number randomized and released (1557) = 195 | 81 of 1557 randomized and released participants Poisson 95% CI = 63–99 | Well below expectation |
| Drug‐related deaths in first 4 weeks and next 8 weeks after release | We expect 1/200 × 1557 = 7.9 or 8 DRDs in first 4 weeks after release if NOR is not effective; and a further 1/800 × 1557 = 1.9 or 2 DRDs in the next 8 weeks |
2 DRDs in first 4 weeks post‐release; a further 3 DRDs in the next 8 weeks were registered with Office for National Statistics by 21 April 2016 First 4 weeks, 95% CI = 0.2–7.2 12 weeks, 95% CI = 1.6–11.7 |
Below expectation for the first 4 weeks; consistent with expectation in the first 12 weeks |
| Non‐fatal overdose‐related admissions within 12 weeks of release | We assume participants’ non‐fatal overdose admissions to Accident and Emergency within 12 weeks of index release to be between two and eight times as many as DRDs with 2–3 times as many DRDs being our best estimate, thus we expect 20–30 (but up to 80) non‐fatal overdose‐related Accident and Emergency admissions | Awaiting Hospital Episode Statistics data from Health and Social Care Information Centre (now NHS Digital) | No information |
CI = confidence interval; DRD = drug‐related deaths; NOR = naloxone‐on‐release.