br We compared patients who died at home to

We compared patients who died at home to those who died in the hospital/facility with respect to the in-dependent variables. Descriptive statistics (frequencies and percentages) were reported for categorical vari-ables and compared using Pearson’s c2 test and Fisher’s exact test where appropriate; means and me-dians were reported for continuous variables and compared using t-tests and Wilcoxon rank-sum tests.
We used binary logistic regression to determine which factors were independently associated with preferred and actual home death. All variables that Lovastatin were significant at P < 0.1 on univariate analysis were evaluated in the multivariate analysis and nonsig-nificant factors were dropped with stepwise backward 
regression. A two-sided P-value of <0.05 was consid-ered significant throughout. We determined the extent of agreement (congruence) between preferred POD and actual POD using Cohen’s kappa coefficient
Results
Table 1 Sociodemographic and Clinical Characteristics of Patients With Advanced Cancer Attending the Palliative Care Unit of CHBAH
Gender
Marital status
Race
Highest level of education
Employment status
informal trading)
Site of primary tumor
hepatobiliary
median (IQR)
Comorbidities
Eastern Cooperative Oncology Group at
baseline (by interviewer)
Eastern Cooperative Oncology Group at
baseline (by caregiver)
Pain medicationsa
Missing data
Patient’s illness understanding
Relatively healthy but terminally ill 4 (2.09)
Time between diagnosis of cancer and
deatha
days, median (IQR)
Time between enrollment and death
(months)a
(Continued) 
Table 1 Continued
Hospital, or other institution (hospice, 64 (33.51) nursing home)
CHBAH ¼ Chris Hani Baragwanath Academic Hospital; IQR ¼ interquartile range; COPD ¼ chronic obstructive pulmonary disease; APCA POS ¼ African Palliative Care Association Palliative Outcome Scale.
aMissing values for the variables are as follows: pain medications (n ¼ 9), time between diagnosis of cancer and death (n ¼ 1), time between enrollment and death (n ¼ 1).
Patients whose caregivers reported that reduction had wanted family and/or friends to be present were more likely to die at home than those who did not or were not sure (OR 6.73, 95% CI 2.97e15.30; P < 0.001). Preference for death at home was not asso-ciated with actual death at home (Table 2).
In a multivariate logistic regression model control-ling for site of primary cancer and preferred POD, dying at home was marginally associated with increasing age (OR 1.04, 95% CI 1.01e1.06; P ¼ 0.01). Patients who wanted family and friends pre-sent were more likely to die at home (OR 7.83, 95% CI: 3.27e18.71). Wealth quintile and preferred POD were not associated with home death (Table 3).
Patients who preferred to die and actually died in hospital were less likely than others (according to the caregiver) to want family or friends to be present (OR 0.14, 95% CI 0.03e0.54, P ¼ 0.001). Patients who preferred to die in hospital but died at home lived
Table 2 Patients’ Characteristics and Associations With Actual and Preferred Home Death
Actual Place of Death
Preferred Place of Death
Hospital/Facility N ¼ 114 Home N ¼ 77 Unadjusted Analysis
Hospital/Facility N ¼ 64 Home N ¼ 127 Unadjusted Analysis
Characteristics
Gender
Marital status
Highest level of education
Wealth quintileb
Site of primary tumor
Comorbidities
Eastern Cooperative Oncology Group at
baseline (by interviewer)
Pain medicationsc
Missing data
Patient’s illness understanding
terminally ill
terminally ill
(Continued) 
Cancer Patients’ Place of Death in South Africa