Part 1 2 3 4 5 6

(vii)      Labor and financial burden of illness on households

72% (182/253) of ill household members were cared for at home, the rest being hospitalized or ambulatory. Indicators of the burden on households are shown in Table 17. Being cared for at home was slightly more likely among those from affected households (75%) than from unaffected households (62%) households. The duration of being cared for at home appeared higher in affected households (median 20) than unaffected households (median 14), but this difference was not significant. Among the 177 for whom the logistical burden of home care was reported, caring for the ill person took a median of 4 hours per day. This took longer in affected households (median 4) than in unaffected households (median 3) (P=0.06). Almost 60 % of ill household members that attended health services was accompanied by someone else when attending these facilities. Those from affected households were significantly more likely (68%) to be accompanied than those from unaffected households (37%).

Table 19: Caring for an ill household member at home

* Exact test for proportions, Wilcoxin ranksum test for medians.

Eight ill household members lost income while ill, all of these coming from affected households. These 8 people lost a median of 13 (range 4-30) days of work due to their illness, resulting in a median loss of income of R220 (range 100-1600, IQR 155-1125) over the previous 30 days.  Caring for an ill person led their caregivers to lose income in 5% (9/180) of cases; this percentage did not differ between affected and unaffected individuals (P=1.0). Among these 9 caregivers, the median number of working days lost over the past month was 7 (range 1-30). Carers came from outside the household in 12/83 (6.5%) cases. Only 5% (7/149) of those accompanying ill household members to health services lost income as a result, and this did not differ between affected and unaffected households (P=1.0).

Mortality among affected households during the previous six months

44 deaths were reported: 42 in affected and 2 in unaffected households (relative risk 21; 95% CI 6-180). Of the 2 deaths in unaffected households, one was a stillbirth and the other was due to dehydration in an infant. Among affected households, 26 deaths occurred among 101 households in QwaQwa, while 16 deaths occurred among 101 households in Welkom (relative risk 1.3; 95% CI 0.65-2.7).

In a multiple logistic regression model including all households, a death was 25 times more likely in affected households, twice as likely in QwaQwa as in Welkom, and about 1.6% more likely with every 1% increase in the percentage of the household that was female (Table 20). The latter figure is equivalent to a 17% increase in risk for a difference of 10% in the percentage that was female (i.e. OR=1.016¹⁰). Income, employment status, medical aid cover and age distribution had no independent influence on risk of death. There was no interaction between affected status and urban/rural location.

Table 20: Predictors of death in a household: logistic regression model*

* Adjusted for household age distribution and number of household members.

The following mortality results will be confined to affected households.

(i)         Demographic characteristics of deceased household members

Among the 42 members who died in affected households, half (49%) were male and half were female. A third (33%) of the deceased were married or living with their partner, while two thirds were unmarried, divorced or widowed. Six (14%) were the former head of the household, 12 (29%) were the informant’s child or stepchild, 7 (17%) their sibling, 2 (5%) their spouse or partner, 5 (12%) their parent, 6 (14%) their grandchild and 4 (10%) a domestic worker.

The mean age of death was 35 (range 0-73, inter-quartile range 24-49) years. The following graph of the age distribution of deaths shows a peak around 35 years, with a smaller peak among young children (Figure 3).

(ii)        Cause of death

The cause of death was reported to be some kind of infectious disease in 33 (79% of 42) cases: tuberculosis in 12 (29%) cases, HIV/AIDS in 12 (29%), pneumonia in 7 (14%) and meningitis in 2 (5%). Thus about four in five deaths were due to infections, which could be HIV-related. Infections accounted for 29 of 33 deaths up to 50 years of age, compared to 5 of 9 deaths over 50 (relative risk 1.6; 95% CI 0.6, 5.2).  The remaining 9 deaths were due to cancer, stroke, diabetes, trauma or unknown causes. 89% (35/40) of the deceased were ill for at least a month before their death.

Table 21: Causes of death

* Excludes 2 with no cause stated.

(iii)       Health care utilization before death

39/41 (95%) sought treatment before death. The commonest source of care was government hospitals, followed by traditional healers, government clinics, and private providers (Table 22).

Table 22: Source of care for fatal illness

* Source not stated for 3 individuals.

(iv)       Labor and financial burden of fatal illness on households

Table 23 shows the frequency distribution of health care costs among households. The mean household cost of health care for the fatal illness (assuming the cost for each person was the midpoint of the respective category) was R167: R56 for consultation fees, R19 for hospital fees, R55 for medicines and R 37 for transport (Table 23). Only one patient had medical aid cover.

Deaths also resulted in a loss of income for households. 6 (14% of 42) of the deceased were reportedly employed. 10 reportedly had an income: 4 under R1000 per month, 5 from R1000 to R2000 per month and 1 over R2000 per month. 3 of 41 households received life insurance or workers’ compensation payments after the death.

Table 23: Cost of health care for fatal illness

# Assumed if cost not reported.

* Total cost estimates assume that the cost for each patient was the midpoint of the respective cost category.

Funeral expenses cost a median of R4000-5000 (inter-quartile range R2000-3000; R6000-7000). The means of finance most often used to pay for funerals were funeral insurance, relatives and/or friends, and own income. Interesting, though, is that a significantly larger percentage of funerals in QwaQwa was paid for with funeral insurance and by friends and relatives compared to Welkom where own income, borrowing and savings was used more often. This makes sense insofar as poverty and unemployment are worse in rural than in urban areas, as will be shown in subsequent discussions.

Table 24: Sources of finance for paying for funeral costs (multiple response)

The logistical burden of caring for the deceased during their fatal illness was as follows. Household members spent an average of 7.5 (range 2-24) hours per days providing care. Loss of income due to caring was however reported for only 2 (5% of 38) households. Care appeared to be provided mainly by unemployed household members: an average of 5 (range 2-10) working days was lost caring for them during the months before their death. Carers were almost always relatives (Table 25).

Table 25. Relationship of carer to the deceased