If information from prognostic studies is to be used by clinicians to derive prognoses of patients early after stroke, it is important that prognostic studies recruit representative populations (Herbert et al 2005) seen early after stroke. These include
consecutive cohorts from hospitals or cohorts from registries, rather than a select group of patients included in trials or referred for rehabilitation. It is also important that studies not only identify significant predictors but develop robust and clinically applicable models Staurosporine mouse for external validation. Without external validation, it is not recommended for clinicians to use the prediction models in clinical practice (Moons et al 2009). Studies that have recruited cohorts early after stroke have reported varying estimates of recovery of independent ambulation (41 to 85%) (Dallas et al 2008, Feigin et al 1996, Veerbeek et al 2011, Wade and Hewer 1987, Wandel et al 2000) and upper limb function (32 to 34%) (Au-Yeung and
Hui-Chan 2009, Heller et al 1987, Nijland et al 2010). In addition, some researchers check details have conducted multivariate analyses of data from acute stroke cohorts. These studies reported that pre-morbid function (Wandel et al 2000), strength of leg muscles (Veerbeek et al 2011, Wandel et al 2000), sitting ability (Loewen and Anderson 1990, Veerbeek et al 2011), walking ability and bowel control (Loewen and Anderson 1990) predicted recovery of independent What is already known on this topic: Many studies have identified predictors of recovery of ambulation and upper limb function after stroke. However, few have recruited representative cohorts early after stroke or developed prediction models suitable for external validation. What this study adds: Within six months of stroke, over two-thirds of people who are initially non-ambulant recover
independent ambulation but less second than half of those who initially lack upper limb function recover it. Prediction models using age and NIHSS can predict independent ambulation and upper limb function six months after stroke. External validation of these models is now required. Two prognostic models, one of ambulation and one of upper limb function, were recently developed by one group in the Netherlands and these are potentially at the stage of external validation (Nijland et al 2010, Veerbeek et al 2011). Even though the cohorts do not appear to have been recruited consecutively, recruitment from multiple acute stroke units and high follow-up rates in both studies may make these cohorts more representative than other non-consecutive cohorts. They also reported good predictive accuracy of their models (positive likelihood ratios = 5.24 to 5.