Data for this study were collected as part of the Asian and African Patient Perspectives Regarding Oncology Awareness, Care, and Health (APPROACH) study, a multi-country cross-sectional study of end-of-life care among advanced cancer patients.

Data for the Kenya site were collected at Moi Teaching and Referral Hospital (MTRH), Eldoret, a level 6 hospital offering specialised oncological and palliative services with a catchment area of 24 million residents in Western Kenya, Eastern Uganda and South Sudan.

Eligible participants were: (1) ≥ 21 years old, (2) diagnosed with Stage IV solid cancer and (3) able to understand English and (4) seeking treatment at MTRH during the study period. Consistent with the multisite APPROACH protocol, we aimed to recruit a sample of 200 participants.^20,21 To obtain this sample size, 473 patients were assessed for eligibility via daily screening of medical records of outpatients seen at the medical oncology and palliative care departments and inpatient medical and surgical wards and referrals from nurses and patient. Of the pool of patients, 266 were deemed ineligible based on medical record review or after being approached for the study or declined to participate. The remaining 207 patients were recruited for face-to-face interviews conducted at MTRH between October 2021 and February 2022. The STROBE participant flow diagram can be found in Figure 1²². The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist can be found in Online Appendix 1.

The survey questionnaire comprised validated scales and items developed by APPROACH study investigators in consultation with oncologists and research faculty at participating centres.²³

We first summarise participants’ characteristics with mean and standard deviations (s.d.) for continuous variables and frequencies and percentages for categorical variables. We then fit two binary logistic regression models to assess the association between participant characteristics and: (1) prognostic awareness and (2) preference for prognostic information. The dependent variables were prognostic awareness (0 = unaware, 1 = aware) and preference for prognostic information (0 = no or not sure; 1 = yes). Independent variables for both models were age, years of education, socioeconomic status and symptom burden.

We also conducted a post hoc estimation of the variance inflation factor (VIF) to assess multicollinearity between the independent variables.²⁵ All analyses were conducted using Stata version 15.1.

Ethics approval was obtained from the National University Singapore-Institutional Review Board (NUS-IRB LB-15-319) and the Moi Institutional Research and Ethics Committee (IREC/2021/27). Trained interviewers obtained written informed consent from all participants prior to the survey. Only the trained interviewers and study team at MTRH had access to participants’ confidential information. Participants were assigned a unique identification number and only de-identified data were collected and analysed.

Table 1 presents sample characteristics. Respondents ranged in age from 21 to 85 with a mean age of 55 (SD = 15.4) and an average of 8.5 (SD = 4.6) years of education. Slightly more respondents were female (57%) than male and most were married (76%). Perceived household economic status varied among participants with 37% indicating low, 38% indicating lower-middle and a quarter indicating upper-middle or high-income. Most participants (53%) had known of their cancer diagnosis for 1 to 3 years. Participants reported a mean symptom burden score of 13.9 (SD = 7.8; range: 0 to 37).

Nearly two-thirds of participants were aware of their prognosis (64%) (Table 1). When we examined participants’ characteristics associated with prognostic awareness in the logistic regression model, we found no significant associations (Table 2).

One-third (33%) of participants indicated a preference to receive prognostic information (Table 1). When we examined participants’ characteristics associated with preference for prognostic information in the logistic regression model, we found significant associations with all participant characteristics (Table 3). Contrary to our hypotheses, increased age was associated with a higher likelihood (odds ratio [OR] = 1.04, 95% confidence interval [CI]: 1.02, 1.07) and increased symptom burden was associated with a lower likelihood (OR = 0.94, CI: 0.90, 0.99) of preference to receive prognostic information. Also contrary to our hypotheses, participants reporting higher perceived household income levels (lower-middle vs. low: OR = 0.19; CI: 0.09, 0.44 and upper middle or high vs. low: OR = 0.22, CI: 0.09, 0.56) were less likely to prefer receiving prognostic information. Supporting our hypothesis, participants reporting higher education levels were more likely (OR: 1.18, CI: 1.08, 1.30) to prefer prognostic information (Table 3). We did not observe multicollinearity in either of the multivariable models (VIF < 2).

The strength of this study lies in its examination of patient prognostic awareness and preference for receiving prognostic information in a country in Africa (Kenya) where little information currently exists. This study has several limitations. One limitation was that the study was conducted in a single site in Kenya, and thus results may not be generalisable to other sites in Kenya or Africa. Likewise, the focus on a single condition (i.e. advanced cancer) and missing data (~5%) are additional limitations. Most importantly, this study evaluates correlations and thus cannot identify causal reasons behind the low levels of prognostic awareness and preference for receiving prognostic information. Future studies should explore causal factors related to low prognostic awareness and information preference among advanced cancer patients in Africa.