MODIFYING THE NUMBER NEEDED TO SCREEN (NNS) TO ACCOUNT FOR THE PREVALENCE DIFFERENCE (PD) OF RISK FACTORS: A NOVEL APPROACH
Author(s)
Alexa C. Bowie, MPH1, CHRISTINA QIAN, MSc1, Lydia Y. Lee, MS, PharmD2, Lindsay Bengtson, PhD2, Jiaming Calvin Liang, MD3, Joshua Reed, MD4, Shelagh Szabo, MSc1.
1Broadstreet HEOR, Vancouver, BC, Canada, 2Boehringer Ingelheim Pharmaceuticals Inc., Ridgefield, CT, USA, 3University of British Columbia, Vancouver, BC, Canada, 4University of Ottawa, Ottawa, ON, Canada.
1Broadstreet HEOR, Vancouver, BC, Canada, 2Boehringer Ingelheim Pharmaceuticals Inc., Ridgefield, CT, USA, 3University of British Columbia, Vancouver, BC, Canada, 4University of Ottawa, Ottawa, ON, Canada.
OBJECTIVES: The NNS is a widely-recognized, easy-to-interpret metric of screening burden and efficiency (NNS to identify one undiagnosed disease case among those with a risk factor or prevent one adverse event). Extending these methods to identify undiagnosed disease indicators can inform targeted screening strategies. While PD may signal potential screening targets through imbalance, it’s less clinically intuitive than the NNS. We present a modified NNS approach accounting for differences in risk factor prevalence.
METHODS: A simulated dataset of hypothetical individuals with ‘undiagnosed’ vs. ‘no-disease’ and risk factors (hypertension, malignancy, diabetes, asthma, and cardiovascular disease) was generated in R. Risk factor prevalence was randomly generated, with higher prevalence assigned to the undiagnosed cohort. PD was computed as the difference in the proportion with the risk factor between undiagnosed vs. no-disease cohorts. NNSmodified was estimated for each risk factor as 1/PD, representing the number with the risk factor needed to screen to identify one additional undiagnosed case, beyond screening those without the risk factor. As an exploratory analysis, a traditional NNS (1/prevalence of undiagnosed disease among a subset with the risk factor) was estimated.
RESULTS: Fifty-thousand simulated patients were generated; n=22,249 ‘undiagnosed’ (44.5%) and n=27,751 ‘no-disease’ (55.5%). Risk factors with greater PD yielded a lower NNSmodified (asthma: 33.1% vs. 6.3%, PD=26.7%, NNSmodified=4), while those with a smaller PD yielded a larger NNSmodified (malignancy: 12.0% vs. 11.6%, PD=0.4%, NNSmodified=238). In comparison, traditional NNS estimates were substantially smaller (e.g., 45.1% of those with malignancy had undiagnosed disease: NNStraditional=3 vs. NNSmodified = 238), highlighting the additional information gained by accounting for PD when prioritizing populations for screening.
CONCLUSIONS: Choice of NNS method depends on the research question. When identifying priority risk factors, the proposed NNSmodified approach provides additional insights to guide targeted screening. Future real-world research to validate these methods will help ensure clinical utility.
METHODS: A simulated dataset of hypothetical individuals with ‘undiagnosed’ vs. ‘no-disease’ and risk factors (hypertension, malignancy, diabetes, asthma, and cardiovascular disease) was generated in R. Risk factor prevalence was randomly generated, with higher prevalence assigned to the undiagnosed cohort. PD was computed as the difference in the proportion with the risk factor between undiagnosed vs. no-disease cohorts. NNSmodified was estimated for each risk factor as 1/PD, representing the number with the risk factor needed to screen to identify one additional undiagnosed case, beyond screening those without the risk factor. As an exploratory analysis, a traditional NNS (1/prevalence of undiagnosed disease among a subset with the risk factor) was estimated.
RESULTS: Fifty-thousand simulated patients were generated; n=22,249 ‘undiagnosed’ (44.5%) and n=27,751 ‘no-disease’ (55.5%). Risk factors with greater PD yielded a lower NNSmodified (asthma: 33.1% vs. 6.3%, PD=26.7%, NNSmodified=4), while those with a smaller PD yielded a larger NNSmodified (malignancy: 12.0% vs. 11.6%, PD=0.4%, NNSmodified=238). In comparison, traditional NNS estimates were substantially smaller (e.g., 45.1% of those with malignancy had undiagnosed disease: NNStraditional=3 vs. NNSmodified = 238), highlighting the additional information gained by accounting for PD when prioritizing populations for screening.
CONCLUSIONS: Choice of NNS method depends on the research question. When identifying priority risk factors, the proposed NNSmodified approach provides additional insights to guide targeted screening. Future real-world research to validate these methods will help ensure clinical utility.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
EPH104
Topic
Epidemiology & Public Health
Disease
No Additional Disease & Conditions/Specialized Treatment Areas