OBJECTIVES: The primary objective was to determine the cost-effectiveness acceptability of targeted (TA) versus universal assessment (UA) and geriatric team (GM) versus primary care (PC) management models. METHODS: The primary effectiveness measure was life years gained (LYGs). The perspective was societal. The primary analysis used a 3-year timeframe. Resource use data and survival were collected within a multi-centre, cluster randomised trial of general practitioner practices. Patients were over 74 years old, living in the community, in the UK. Costs and outcomes were discounted (3.5% recommended UK rate). Missing data for censored cases were imputed by survival analysis. Missing data due to missing observations were imputed by characteristic of patient. Data were adjusted for age, gender and cluster randomisation. Costs and LYGs were bootstrapped. Net benefit statistics were estimated. Cost-effectiveness acceptability analysis used willingness to pay thresholds (GBP0 to GBP50000). Sensitivity analysis assessed the impact of structural factors and assumptions. RESULTS: 109 GP practices were assigned to (a) assessment method: UA=55 (21762 patients) TA=54 (21457 patients); (b) management method: GM=55 (22216 patients); PC=54 (21003 patients). Preliminary analysis indicated a net cost to TA (GBP296; 2.5-97.5 percentile GBP140-GBP448 ) versus UA and a net saving to PC (-GBP41; 2.5-97.5 percentile -GBP192-GBP107) versus GM. LYGs were TA (0.006; 2.5-97.5 percentile –0.006-0.19) and PC (0.016; 2.5-97.5 percentile 0.004-0.28). The probability of net benefit was 0-0.50 for TA across the willingness to pay thresholds. The probability of net benefit for PC was 0.70-1.00. Sensitivity analysis indicated the results for targeted assessment, but not PC, were sensitive to method of imputing missing data and timeframe. CONCLUSIONS: The cost-effectiveness of targeted assessment is uncertain. PC management appears cost effective in the primary and sensitivity analyses. The small cost and effect differences between strategies indicate cost-effective configuration of services may be driven primarily by local considerations.