Data envelopment analysis to evaluate the efficiency of tobacco treatment programs in the NCI Moonshot Cancer Center Cessation Initiative

Abstract

Abstract Background The Cancer Center Cessation Initiative (C3I) is a National Cancer Institute (NCI) Cancer Moonshot Program that supports NCI-designated cancer centers developing tobacco treatment programs for oncology patients who smoke. C3I-funded centers implement evidence-based programs that offer various smoking cessation treatment components (e.g., counseling, Quitline referrals, access to medications). While evaluation of implementation outcomes in C3I is guided by evaluation of reach and effectiveness (via RE-AIM), little is known about technical efficiency—i.e., how inputs (e.g., program costs, staff time) influence implementation outcomes (e.g., reach, effectiveness). This study demonstrates the application of data envelopment analysis (DEA) as an implementation science tool to evaluate technical efficiency of C3I programs and advance prioritization of implementation resources. Methods DEA is a linear programming technique widely used in economics and engineering for assessing relative performance of production units. Using data from 16 C3I-funded centers reported in 2020, we applied input-oriented DEA to model technical efficiency (i.e., proportion of observed outcomes to benchmarked outcomes for given input levels). The primary models used the constant returns-to-scale specification and featured cost-per-participant, total full-time equivalent (FTE) effort, and tobacco treatment specialist effort as model inputs and reach and effectiveness (quit rates) as outcomes. Results In the DEA model featuring cost-per-participant (input) and reach/effectiveness (outcomes), average constant returns-to-scale technical efficiency was 25.66 (SD = 24.56). When stratified by program characteristics, technical efficiency was higher among programs in cohort 1 (M = 29.15, SD = 28.65, n = 11) vs. cohort 2 (M = 17.99, SD = 10.16, n = 5), with point-of-care (M = 33.90, SD = 28.63, n = 9) vs. no point-of-care services (M = 15.59, SD = 14.31, n = 7), larger (M = 33.63, SD = 30.38, n = 8) vs. smaller center size (M = 17.70, SD = 15.00, n = 8), and higher (M = 29.65, SD = 30.99, n = 8) vs. lower smoking prevalence (M = 21.67, SD = 17.21, n = 8). Conclusion Most C3I programs assessed were technically inefficient relative to the most efficient center benchmark and may be improved by optimizing the use of inputs (e.g., cost-per-participant) relative to program outcomes (e.g., reach, effectiveness). This study demonstrates the appropriateness and feasibility of using DEA to evaluate the relative performance of evidence-based programs.