Article Text

Download PDFPDF
The Impact of Sociodemographic Factors and Emergency Admissions on the Place of Death of Gynecological Cancer Patients in England: An Analysis of a National Mortality–Hospital Episode Statistics–Linked Data Set
  1. Sonali Kaushik, MD, MRCOG*,
  2. Luke Hounsome, PhD,
  3. Catherine Blinman, MRCP,
  4. Robert Gornall, DM, MRCOG and
  5. Julia Verne, PhD§
  1. *Royal Sussex County Hospital, Brighton;
  2. Public Health England National Cancer Registration and Analysis Service, Bristol;
  3. Cheltenham General Hospital, Cheltenham;
  4. §End of Life Care Public Health England, Bristol, England.
  1. Address correspondence and reprint requests to Sonali Kaushik, MD, MRCOG, Division of Gynaecological Oncology, Royal Sussex County Hospital, Eastern Road, Brighton, England BN2 5BE. E-mail: Sonali.kaushik{at}


Objective The aim of this study was to develop a predictive model for risk of death in hospital for gynecological cancer patients specifically examining the impact of sociodemographic factors and emergency admissions to inform patient choice in place of death.

Methods The model was based on data from 71,269 women with gynecological cancer as underlying cause of death in England, January 1, 2000, to July 1, 2012, in a national Hospital Episode Statistics–Office for National Statistics database. Two thousand eight hundred eight deaths were used for validation of the model. Logistic regression identified independent predictors of a hospital death: adjusting for year of death, age group, income deprivation quintile, Strategic Health Authority, gynecological cancer site, and number of elective and emergency hospital admissions and respective total durations of stay.

Results Forty-three percent of deaths from gynecological cancer occurred in hospital. The variables significantly predicting death in hospital were less recent year of death (odds ratio [OR], 0.93; P < 0.001), increasing age (OR, 1.17; P < 0.001), increasing deprivation (OR, 1.06; P < 0. 001), increasing frequency and length of elective and emergency admissions (P < 0.001). The model correctly identified 73% of hospital deaths with a sensitivity of 75% and a specificity of 72%. The areas under the receiver operating curve were 0.78 for the predictive model and 0.71 for the validation data set. Each subsequent emergency admission in the last month of life increased the odds of death in hospital by 2.4 times (OR, 2.38; P < 0.001). Hospital deaths were significantly lower in all other regions compared with London. The model predicted a 16% reduction of deaths in hospital if 50% of emergency hospital admissions in the last month of life could be avoided by better community care.

Conclusions Our findings could enable identification of patients at risk of dying in hospital to ensure greater patient choice for place of death.

  • End of life care
  • Place of death
  • Trends
  • Hospital death
  • Emergency admissions
  • Gynecological cancer

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Author Contributions: S.K., C.B., J.V., and R.G. devised concept of study and study design. S.K., L.H., and J.V. collected data and were jointly responsible for data preprocessing, development of the prediction model, data analysis, and literature search. S.K. and L.H. constructed the tables. All authors contributed to the writing of the manuscript and interpretation of data.

    Public Heath England maintains and manages the national person-based, linked Hospital Episode Statistics–Office for National Statistics database. Our study is an exploratory analysis of these data. Any external funding or sponsorship does not support the analysis. S.K., L.H., and J.V. had access to raw data, and S.K. had final responsibility to submit for publication.

    No external funding has been received for this study.

  • The authors declare no conflicts of interest.

    All authors have completed the ICMJE uniform disclosure form at

  • No ethical approval required. All patient data used in this study were handled and processed in accordance with National Health Service best practice and Caldicott recommendations. The Data Protection Act does not specifically apply to deceased patient records; any common law duty was adhered to.

    Supplemental digital content is available for this article. Direct URL citation appears in the printed text and is provided in the HTML and PDF versions of this article on the journal’s Web site (