Cover Image
close this bookDistrict-based Health Insurance in sub-Saharan Africa (Institut Tropical - Tropical Institute, Antwerp, Belgium, 1998)
close this folder2. The impact of the Bwamanda scheme on hospital utilisation patterns
View the documentIntroduction
View the documentMethods and sources of data
View the documentResults
View the documentDiscussion

Results

DISTRIBUTION OF THE HIGHER HOSPITAL UTILISATION AMONG INSURED OVER THE DIFFERENT WARDS

These data were already presented in the previous chapter (Table 5 and Table 6). In the period April 1993 - March 1994, 7,362 hospital admissions took place. Fifteen per cent (1,078/7,362) of all admissions occurred for people living outside the district. The overall admission rate, for the entire district population, is 39.2‰ (6,284/160,267). The admission rate for insured was nearly 3 times higher than for non-insured. The ward specific ratios of insured: non-insured admission rates were highest for surgery (where the admission rate for insured is more than 10 times the one for non- insured) and for the maternity (a sevenfold admission rate for insured). These ratios were of a same order of magnitude in paediatrics and gynaecology (respectively 3.8 and 3.4), but much smaller in the departments of intensive care and internal medicine (respectively 1.43 and 1.35).

SPATIAL ANALYSIS

Figure 9 shows the total admission rates per health centre area, broken down by insurance status. In all health centre areas, the utilisation is higher for insured than for non-insured (Wilcoxon signed ranks test significant, P<0.001). Both for insured and for non-insured there is a clear distance decay, although more pronounced for the insured.

The higher utilisation by insured appears to be concentrated mainly in the populations living in a health centre area within a range of about 35 km from the hospital. Figure 10 suggests the existence of two clusters. A first is made up by 5 of the 17 health centre areas, where the utilisation differentials (the difference between the rates for the insured and non-insured) range between 18 and 35‰ admissions per year. These areas are less than 35 km from the hospital.

A second cluster concerns 12 health centre areas where the differential is between less than 1 to about 17 admissions per 1000 inhabitants (with a majority at approximately 5 or 6 admissions per 1000 inhabitants). These centres are located between 35 and 100 km from the hospital.


Figure 9: Admission rates per health centre area


Figure 10: Utilization differentials insured/non-insured

In 15 of the 17 health centre areas insurance is associated with higher hospital admission rates (Figure 11), with little influence of distance. For the health centres within a range of 35 km from the hospital, the average ratio admission rate insured: non-insured is 6; for the 12 at more than 35 km from the hospital, the ratio is 4.5. When the four outliners with individual ratios above 8 are excluded, these ratios are respectively 3.5 and 2.9.


Figure 11: Ratios per health centre of admission rate insured/non-insured


Figure 12: Utilization differential per hospital department

These differences remain after dis aggregation of the data for surgery, internal medicine/gynaecology, and paediatric departments (Figure 12). Utilisation is significantly higher for insured than for non-insured (Wilcoxon signed ranks test significant at the level 0.001 for surgery and paediatrics, and at the level 0.05 for internal medicine/gynaecology).

Table 8. Caesarean section and symphysiotomy rates in insured and non-insured populations for 5-year period April 1991 - March 1996.

Health centres

N expected pregnancies among insured

N of intervention

Rate per 100 insured

N expected pregnancies non-insured

N of interventions

Rate per 100 non-insured

Botuzu

1048

24

2.29

487

3

0.62

Kada

1002

24

2.40

549

9

1.64

Botela

1462

26

1.78

901

9

1.00

Bombese

414

5

1.21

688

3

0.44

Bongbada

1255

19

1.51

692

3

0.43

Boto

699

18

2.57

1337

9

0.67

Bowazi

641

9

1.40

215

6

2.79

Bodeme

1372

29

2.11

1275

9

0.71

Kasongo

267

6

2.25

1018

5

0.49

Bowara

862

24

2.78

590

6

1.02

Bogbase

571

5

0.88

309

4

1.30

Bobisi

506

5

0.99

305

2

0.66

Bobandu

310

6

1.93

1178

2

0.17

Bolumba

422

11

2.61

636

8

1.26

Bombisa

927

22

2.37

355

4

1.13

Bowakara

250

5

2.00

455

0

0.00

Bokozo

56

0

0.00

361

2

0.55

Total

12,064

238

1.97

11,351

84

0.74

NOTES: health centres are listed according to increasing distance from the hospital; expected births are calculated assuming a birth rate of 40 per 1000.


Figure 13: Admission rates caesarean sections and symphysiotomies

Table 8 shows the rates of caesarean sections and symphysiotomies per 100 expected deliveries in the 1991-1996. In this five-year period, 238 caesarean sections/symphysiotomies11 were carried out in the insured population out of 12,064 expected deliveries (1.97%), and 84 out of 11,351 expected deliveries in the non-insured population (0.74%). In 14 of the 17 health centre areas the caesarean section rate is higher in the insured population. In the 3-year period 1993-1996, the proportion of interventions carried out for absolute maternal indications in both insured and non-insured populations was identical (93%).

11 Approximately one symphysiotomy for every ten caesarean sections

The caesarean section rate in the insured population appears to be relatively independent of distance, whereas these rates decrease significantly with distance in the non-insured population (Wilcoxon signed ranks test, P<0.001) (Figure 13).

In the 3-year period 1993-1996, 119 insured patients (80% males) underwent surgery for a strangulated hernia (6.41 per 10,000 inhabitants per year). In 2 out of the 17 health centres areas under investigation, no admissions for strangulated hernias took place in the insured population in the three-year time-span. In the non-insured population, this annual rate was 2 per 10,000 inhabitants (35 patients of which 74% males), with no admissions coming from 5 of the 17 health centre areas. The difference between these two rates is significant (Wilcoxon signed ranks test, P=0.01).

The average utilisation differentials for strangulated hernias are roughly similar for the two groups of health centres located at less than 35 km from the hospital and beyond 35 km (Figure 14). In the first case, the average differential is 4.3/10,000, and in the second case, it is 3.1/10,000

The data for uncomplicated hernias indicate that a total of 597 insured patients (68% male patients), coming from all of the 17 health centre areas, had surgery in the period 1993-1996 (32 1 per 10,000 per year). There were 50 non-insured patients (74% male patients) coming from 16 out of 17 health centre areas (2.9 per 10,000 per year). The difference between insured and non-insured is statistically significant (Wilcoxon signed ranks test: P=0.001). The average utilisation differential is however more pronounced for the nearby communities than for the more remote ones (respectively 40/10,000 and 24/10,000)


Figure 14: Utilization differentials hernias

SPATIAL CONCENTRATION: COEFFICIENTS OF LOCALISATION AND LOCATION QUOTIENTS

Coefficients of localisation, for insured and non-insured, were calculated for the hospital admissions in the paediatric, internal medicine/gynaecology and surgery wards; as well as for the admissions that took place for uncomplicated hernias, strangulated hernias and caesarean sections & symphysiotomies. The results are presented in Table 9. These coefficients of localisation are systematically higher among the non-insured patient populations. The ratios of coefficients of localisation non-insured: insured are highest for caesarean sections & symphysiotomies and strangulated hernias, and lowest for paediatrics and internal medicine/gynaecology.

The location quotients for insured and non-insured were calculated for the admissions for caesarean sections and symphysiotomies, uncomplicated hernias, and strangulated hernias, coming from each of the 17 health centre areas. There are no significant differences between insured and non-insured, but Figure 15 clearly shows that the spread of LQi is systematically more narrow for the insured populations.

Table 9: Coefficients of localisation


CL insured

CL non-insured

Ratio CL non-insured/insured

Paediatricsa

0.32

0.37

1.2

Internal medicine and gynaecologya

0.27

0.39

1.4

Surgery (M&F, including hernias)a

0.23

0.33

1.4

Uncomplicated hernias (M&F)b

0.23

0.4

1.7

Strangulated hernias (M&F)b

0.23

0.43

1.9

Caesarean sections/symphysiotomiesc

0.12

0.23

1.9

a data for 1993-1994; b data for 1993-1996; c data for 1991-1996


Figure 15: Location quotients for insured (I) and non-insured (NI) admissions

AVERAGE LENGTHS OF STAY

Insured patients stay an average of 10.5 days (1,417 admissions/14,885 hospital days), the uninsured 9 days (168 admissions/1,509 hospital days) (Wilcoxon rank sum test not significant).

THE BED CENSUS

The bed census results were analysed for the inpatients living in the district: 170 out of the total of 218 patients present at the moment of the census. Bed occupation was deemed not appropriate in 58/139 insured patients (42%), in 3/9 self-employed non-insured patients (33%), and in 12/22 patients covered by employer-organised schemes (55%). For the total populations of 170 inpatients, this proportion is 43% (73/170). In half of these cases, the reason was that ''patients were waiting to be taken home'. The results of the subjective evaluation of the justification of the admission indicate a very high proportion of justified admissions in both insured and non-insured In 133 out of 139 insured cases (96%) and in 8/9 of the non-insured self-employed cases (89%), the admission was categorised as justified or as avoidable only on condition of substantial changes at health centre level It is however not possible to draw any statistical inference from any of these data, given the small sample of non-insured in the inpatient population.