Where the Threshold Sits: Prolonged Grief in Kenya, Namibia and South Africa
- Among 1,730 bereaved young adults in Kenya (n = 629), Namibia (n = 586) and South Africa (n = 515), probable ICD-11 prolonged grief disorder under the strict scoring algorithm was 9.63% in Kenya, 6.85% in South Africa and 5.34% in Namibia, close to the global benchmark of 9.8% the authors cite.
- The identical responses produced 21.0% to 24.5% caseness under the moderate algorithm and 7.07% to 10.3% under the Maciejewski algorithm. The scoring rule moved the rate up to 4.6-fold while the sample stayed the same.
- Loss exposure was heavy and did not translate into elevated disorder rates: 74.4% had lost two or more people, 47.7% three or more, and 31.8% reported trauma exposure. Violent death accounted for 11.9% of losses in South Africa and 20.3% in Namibia.
- Only closeness to the deceased replicated across all three countries (unstandardised coefficient for a distant versus close relationship: Kenya -6.02, Namibia -4.66, South Africa -5.59, all p < .001); violent or unexpected death was significant in Kenya (+3.73, p < .01) and South Africa (+4.09, p < .05) but not in Namibia. The full risk models for the standard scale explained 6.0% to 11.2% of the variance in grief severity.
ICD-11 admitted prolonged grief disorder with a clause almost no clinician can operationalise: the reaction must exceed what is normal for the person's own social, cultural or religious context. Nobody has written down what those norms are. The authors describe their work as one of the preliminary studies to run the ICD-11 grief criteria and the International Prolonged Grief Disorder Scale (IPGDS) concurrently in three African countries, and what it exposes is less about African grief than about how unstable our own thresholds are once you move them out of the settings where they were built.
Same responses, three prevalence rates
The data come from Wave 4 of the Africa Long Life Study, a longitudinal cohort of young adults recruited from urban and rural communities in 2022 and early 2023, with a pre-registered analysis plan. Of 3,027 respondents, 1,730 reported the death of a loved one and completed the IPGDS: 58.7% women, mean age 19.8 years (SD 0.95).
The instrument held up. Internal consistency for the 14-item standard scale was .93 in the pooled sample (.91 to .94 by country), and the six culturally specific supplementary items reached alpha .87. A single factor was retained by design rather than forced by the data: all item loadings cleared .40, but the second eigenvalue sat only slightly above one in every country and in the pooled sample. Convergent correlations, however, came in lower than predicted: depression r = .20, anxiety r = .24, post-traumatic stress r = .27.
Then the thresholds. Under the strict algorithm, probable PGD ran 9.63% (Kenya), 6.85% (South Africa), 5.34% (Namibia). Under the Maciejewski rule the same answer sheets gave 10.3%, 8.87% and 7.07%. Under the moderate algorithm they gave 23.8%, 21.0% and 24.5%. Nothing changed but the cut-point. Worth noting that the paper itself quotes two different global reference figures, 9.8% in the abstract and 13% in the discussion, which tells you how unsettled the comparator is.
Sex differences ran in the expected direction but not uniformly: under the strict algorithm 12.5% of Kenyan women met criteria against 6.94% of Kenyan men, while in Namibia the gap all but vanished (5.45% versus 5.22%). Those bereaved less than six months ago scored higher than those bereaved longer (mean 34.8 versus 29.9, p < .001), which is what the criterion set predicts and a reassuring sign the scale is tracking something real.
Against that, the exposure data make the modest strict-algorithm rates the interesting result: three quarters of this sample had buried more than one person before their twentieth birthday, and the disorder rate still sat within the international range the authors line up for comparison, which runs from 1.3% in Ireland to 13.5% in China.
What survives the trip across a border
One predictor held everywhere: how close the person was to the deceased. Violent or unexpected death was significant in Kenya and South Africa but not in Namibia. Everything else was local too. Being male predicted lower grief severity in Kenya and Namibia but not South Africa. Severe financial difficulty was a significant risk factor in Kenya and South Africa but not Namibia. The authors are careful not to read the low rates as proof of cultural resilience: under-recognition and a Western-built instrument are equally live explanations.
Three things follow for practice. First, when you report a screening result, report the algorithm with it. A figure of 7% and a figure of 23% here describe the same people. Second, the standard-scale risk models explained at most 11.2% of variance in symptom severity, so a risk profile is a prompt for a conversation, not a substitute for one. Third, watch what the culturally specific items did. The most strongly endorsed across all three countries was "I constantly look back upon the past relationship"; the least endorsed was "shattered my trust in life or faith in God/a higher spiritual power". Preoccupation was intense while trust in life and the spiritual frame stayed intact. If you treat a patient's religious practice as avoidance by default, you may be pathologising the part of the system that is still working.
A prevalence figure for prolonged grief describes the scoring rule at least as much as it describes the population it was measured in.
Cross-sectional self-report from a single wave in a narrow age band of students and school-leavers around 19 to 20 years old, so the rates should not be read as national prevalence for older bereaved adults; the questionnaires were administered in English rather than participants' first languages, no second validated grief measure was included for convergent validity, and the authors did not conduct formal measurement invariance testing across the three countries.