Table 2. Full-sample cost-effectiveness analysis (mean cost in £ / mean QALYs, 95% CI).
| MMHU (309 patients) | Standard care a (290 patients) | Incremental cost / QALYs gained | |
|---|---|---|---|
| The cost of care b c | 7266 (6707, 7861) | 7956 (7307, 8681) | -690 (-1571, 246) |
| The cost of care–adjusted d | 7345 (7248, 7441) | 7862 (7758, 7965) | -517 (-660, -374) |
| Additional MMHU cost | 368 (334, 410) | 0 | 368 (334, 410) |
| Total cost(care cost + MMHU cost) | 7634 (7062, 8253) | 7956 (7307, 8681) | -322 (-1219, 621) |
| Total cost- adjusted (care cost adjusted+ MMHU cost) | 7714 (7606, 7822) | 7862 (7758, 7965) | -149 (-298, 4) |
| QALYs gained b | 0.111 (0.101, 0.121) | 0.103 (0.093, 0.114) | 0.008 (-0.005, 0.020) |
| QALYs gained–adjusted e | 0.109 (0.102, 0.116) | 0.108 (0.101, 0.114) | 0.001 (-0.006, 0.008) |
| ICER | MMHU dominant | ||
| ICER adjusted | MMHU dominant |
aGeriatric ward (204 patients) and general ward (86 patients).
bPrimary care cost and QALY imputed using Multiple imputation by chained equation (MICE). Multiple imputation model applying predictive mean matching (pmm) for costs and utilities, and ordered logit (ologit) for Barthel ADL scores, DEMQOL, and NPI; 100 imputations generated.
cHealthcare (inpatient, day-case, outpatient, EMAS, MHT, critical care, primary care) and social care cost.
dAdjusted by age, sex, utility and permanent care home residence at baseline, and one year pre-admission healthcare cost care cost. A GLM model (family–gamma, link–log) was applied, as it was found to be optimal upon diagnostic procedure on each imputation (the worst test results across imputations were: Park test for gamma family, p-value = 0.05, Pregibon link test, p-value = 0.36, Hosmer-Lemeshow test, p-value = 0.11).
eAdjusted by age, sex, and baseline utility, permanent care home residence, number of medical conditions, NPI, and Barthel ADL. A GLM model (family–normal, link–power 0.25) was applied, as it was found to be optimal upon diagnostic procedure on each imputation (the worst test results across imputations were: Park test for normal family, p-value = 0.02, Pregibon link test, p-value = 0.50, Hosmer-Lemeshow test, p-value = 0.07, with Park test p-value being higher than 0.05 for 95% imputations and with average Park test p-value across imputations equal to 0.41).