• 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • br The exploration of socioeconomic and demographic fac tors


    The exploration of socioeconomic and demographic fac-tors focused on 104 patients in the external validation set. A total of 62 (60%) had next of kin staying at their address. There was no significant difference in mean length of stay when the group that stayed less than 15 days was tested (test statistic t 1.04, p = 3), but it GW 311616 approached significance when a stay of up to 50 days was investigated (without next of kin: mean 9.2 days, with next of kin 6.27 days, test statistic t - 1.76, p = 0.08). In the group that stayed less than 50 days, the distance from home to the hospital did not correlate with an increased duration of stay (r = 0.004, p = 0.9). Finally, Index of Multiple Deprivation deciles were not associated with a longer stay (r = -0.05, p = 0.63). r> D. Tighe et al. / British Journal of Oral and Maxillofacial Surgery xxx (2019) xxx–xxx 5
    Fig. 2. Decision tree to identify long compared with short hospital stays.
    Fig. 3. Histogram to compare frequency of observed with predicted length of stay by type.
    Modelling for the complexity of cases allows for a more meaningful interpretation of data on length of stay (and all outcome data). We found significant differences in case mix between units, which underline the importance of risk adjustment. Linear regression and decision tree models that
    have established important predictors of increased length of stay are age, alcohol intake, performance status, and T stage on presentation, together with free tissue transfer and tracheostomy.
    Raw data suggested considerable differences between the units with the shortest (Site 1) and the longest durations of stay (Site 3, Site 4). These persisted when case-mix adjust-
    6 D. Tighe et al. / British Journal of Oral and Maxillofacial Surgery xxx (2019) xxx–xxx
    Fig. 4. Scatter diagram to compare length of stay (observed compared with predicted) by treatment centre (QVH = Queen Victoria Hospital; RMH = Royal Marsden Hospital; EKHUFT = East Kent Hospitals University Foundation Trust).
    ment was taken into account in those with stays of less than15 days.
    The modelling of extreme lengths of hospital stay (or outliers) was not the focus of this paper, though for pur-poses of allocating resources and hospital finances, blastocyst will be useful to explore further.3,4 This, however, is likely to be difficult if preoperative data are used alone without the incorporation of data on postoperative events, which we deliberately avoided doing in this paper. Further efforts to model for stays of over 15 days, and certainly for those over 50 days, could include data on postoperative complications, because they affect 50% to 70% of patients. Complication rates could be published separately, perhaps to emphasise severity.2,5
    These “length of stay” models for this group of patients are new in the UK. The American College of Surgeons’ National Surgical Quality Improvement Program (NSQIP) risk calculator is an alternative model that predicts length of stay based on preoperative data. A recent audit to validate it using retrospective multicentre data in an Australian group (n = 127) of patients who had glossectomy found that it sig-nificantly underestimated length of stay.6 However, they also cited potential confounding factors that were not taken into account by the risk calculator: age, sex, rural or urban loca-tion, TNM classification, previous chemoradiotherapy, neck dissection, free flap reconstruction, and grade of surgical risk. Another study, which used the NSQIP risk calculator for a group of 157 patients who had ablative surgery on the head and neck with microvasculature reconstruction with fibular flaps, reported only slight concordance between the observed and actual lengths of stay.7 Despite some studies highlight-ing the role of postoperative complications in longer stays,
    we think that these data are not helpful when the length of stay is used as a marker of quality of care.8,9
    To effectively model postoperative outcomes, we assessed the socioeconomic and demographic factors that may be implicated. As we explored this possibility on the last group only, the power of the analysis is limited, but the data sug-gest that patients who live with a member of the family may need to spend less time in hospital. There was no correlation with a standard index of socioeconomic deprivation (IMD decile) or distance from the hospital. Further work will focus on how often a change in social circumstances is needed post-operatively, as this could function as a proxy marker of the quality of surgical care. This question will be included in the collection of further prospective data.