Evaluating Evidence on Intermediate Outcomes: Considerations for Groups Making Healthcare Recommendations

Evaluating Evidence on Intermediate Outcomes: Considerations for Groups Making Healthcare Recommendations (Article in Advancing the U.S. Preventive Services Task Force Methods: Important Considerations in Making Evidence-Based Guidelines, Special Issue of The American Journal of Preventive Medicine).
Jonas, Daniel E. Ferrari, Renee M. Wines, Roberta C. Vuong, Kim T. Cotter, Anne. Harris, Russell P.
2018
American Journal of Preventive Medicine
54(1 sp. 1)S38-S52
https://ac.els-cdn.com/S0749379717304889/1-s2.0-S0749379717304889-main.pdf?_tid=spdf-3cee4f1d-2471-4c61-96a0-f610ac169546&acdnat=1519913485_48c2f0edc95cb9ffebfb583c43bf9274

Abstract:

Introduction

Groups making recommendations need evidence about whether preventive services improve health outcomes (HOs). When such evidence is not available, groups may choose to evaluate evidence about effects on intermediate outcomes (IOs) and the link between IOs and HOs. This paper aims to describe considerations for assessing the evidence linking changes in IOs to changes in HOs.

Methods

Working definitions of IOs, HOs, and other outcomes were developed. All current U.S. Preventive Services Task Force (USPSTF) recommendations through April 2016 were examined to identify how evidence of the IO–HO link was gathered and the criteria that appeared to be used to determine the adequacy of the evidence. Methods of other expert and recommendation-making groups were also examined.

Results

Forty-four USPSTF recommendations involved a relevant IO–HO link. The approaches used most commonly to gather evidence about the link were selected review (19 of 44, 43%) and systematic review (12 of 44, 27%). Some key considerations when assessing the adequacy of evidence about the IO–HO link include adjustment for confounding, proximity of the IO to the HO in the causal pathway, and independence of IO–HO relationship from specific treatments.

Conclusions

Considerations were identified for recommendation-making groups to use when gathering and assessing the adequacy of evidence about the IO–HO link. Using a standard set of written principles could improve the transparency of assessments of the IO–HO link, especially if used together with judgment in a reasoned conjecture and refutation process. Ideally, the process would result in an estimate of the magnitude of change in HOs that is expected for specified changes in IOs.