Modelos prognósticos para predizer desfechos clínicos de pessoas com dor lombar submetidos a tratamentos conservadores: uma revisão sistemática

Abstract

Introduction: One of the greatest contributors to the impact of low back pain on health care systems is the amount of people receiving low-value care. This also includes those people experiencing poor treatment outcomes because they may look for alternative treatments and, consequently, increase the costs related to low back pain. Prognostic models can help identify those people experiencing poor or successful outcomes after receiving a treatment. Objective: To identify the evidence around the available prognostic models to predict successful/poor health outcomes in people with low back disorders receiving conservative treatment. Methods: This review was prospectively registered in PROSPERO (CRD42022370503). Literature searches were conducted in three electronic databases from their inception. Inclusion criteria: studies recruiting adults (aged more than 18 years old) with low back disorders reporting back pain and/or radiating leg pain, who received conservative treatment; studies developing and/or validating prognostic models; study designs that conducted population follow-up, such as prospective cohort studies, randomized controlled trials or register-based studies. Exclusion criteria: studies investigating individual prognostic factors; studies without conservative intervention; studies investigating low back pain related to serious pathology (e.g. infection, tumors, fractures). Two independent reviewers performed the study selection as well as the data extraction and risk of bias assessment of included studies. The main findings of the included studies were narratively described and the discrimination measures of the prognostic models were graphically displayed in a forest plot. Results. Seventy-four studies were included in this review. Most prognostic models contained a set of predictors including socio-demographic variables, clinical features, and self-reported measures. Regarding included studies testing the internal validity of prognostic models predicting pain intensity, nine models showed an inadequate discriminatory capacity (area under the curve (AUC) lower than 0.7), four models showed an adequate discriminatory capacity (i.e., AUC between 0.7 and 0.8), and two models showed an excellent discriminatory capacity (i.e., AUC > 0.8). For disability, three models showed an inadequate discriminatory capacity, three models showed an adequate discriminatory capacity,

and one model showed an excellent discriminatory capacity. The calibration measures of the internal validity of the prognostic models predicting pain intensity and disability were adequate. Regarding external validity, only one model provided both discrimination and calibration measures. Although the calibration measures of this model showed to be adequate, the discrimination measure was inadequate. Conclusion: Several models have been developed to predict pain intensity and disability levels of people with low back disorders receiving conservative treatment. However, only a few were externally validated and future studies should focus on testing it before implementation in clinical practice.