|| Checking for direct PDF access through Ovid
To investigate the validity and reliability of accelerometers to detect lying, sitting and standing postures or purposeful activity in hospitalized adults recovering from acute or critical illness.CINAHL, MEDLINE, EMBASE, AMED, Cochrane Library, PEDro, PsycINFO and SPORTDiscuss were searched from inception to June 2017. Professional networks and reference lists of relevant articles were also searched. The main selection criteria were hospitalized adults with acute or critical illness and studies investigating the validity or reliability of accelerometers to identify body position or purposeful activity.Two authors individually assessed study eligibility and independently undertook methodological quality assessment and data extraction from selected articles. A narrative synthesis of the data was undertaken.Fifteen studies, collectively enrolling 385 hospitalized participants, were identified. Populations included stroke, the elderly, acute exacerbation of chronic respiratory disease, abdominal surgery and those recovering from critical illness. Correlations of r = 0.36 to 0.98 and levels of agreement of κ = 0.28 to 0.98 were reported for identification of lying, sitting or standing postures. Correlations of r = 0.4 to 0.8 with general activity were found, with r = 0.94 and 0.96 reported for step count. The reliability of accelerometry measurement was investigated in one study evaluating step count quantification (intraclass correlation coefficient (ICC) = 0.99, 95% confidence interval (CI) = 0.99–1.00).The validity of accelerometers to determine lying, sitting and standing postures or quantify purposeful activity within hospitalized acute or critically ill populations is variable. The reliability of accelerometry measurement within this setting remains largely unexplored.