Introduction: Prognostic factors can potentially be used to target groups at most risk and improve care received by patients. However, producing reliable and informative systematic reviews (SR) synthesizing evidence regarding potential prognostic factors for a particular outcome can be problematic.
Objectives: To present the experience of conducting three prognostic factor SRs and discuss challenges encountered.
Methods: The three reviews:
• Predictors of unplanned hospital admissions, mortality, and poor quality of life in older people who are identified as frail in primary care.
• Prognostic factors and models for predicting work absence in adults with musculoskeletal conditions.
• Prognostic factors and inequalities in poor outcome in people with dementia. All SRs were conducted using standard, rigorous SR methodology in accordance with guidance from the Cochrane Prognosis group and the Cochrane Handbook.
Results: Inconsistent labelling and reporting of prognosis research required sensitive search strategies to ensure all eligible studies were retrieved. Prognostic research search filters did not substantially reduce the number of papers. Extra screening stages were needed owing to the volume of articles identified: initial title screening by one reviewer (sample by a second reviewer), before abstracts and full texts were independently screened by two reviewers. Variation in reporting required a complex spreadsheet and contributed to difficulties extracting consistent data and assessing risk of bias. Piloting eligibility criteria and spreadsheets was essential to improve reliability. Individual prognostic factors identified were grouped. However, often insufficient, inconsistent data and too much heterogeneity was present for meta-analyses to be conducted. Narrative synthesis assessed potential of prognostic factor groups to influence each outcome and level of confidence (informed by GRADE). Patient and Public Involvement & Engagement (PPIE): PPIE groups proposed potential prognostic factors which differed from the factors found in the published research.
Conclusions: Reviewers encountered several challenges throughout the review process. This was mainly due to variability in conduct and reporting of studies and the large numbers of studies and individual prognostic factors identified and screened in each of the SRs. The experiences of the reviewers should be considered with reference to new research methods, such as automation, and their potential for solving these challenges.