Community-based health insurance and healthcare service utilisation, North-West, Ethiopia: a comparative, cross-sectional study

Abstract

Objectives

The objective of this study was to compare differences in healthcare utilisation between community-based health insurance member households and non-member households and to identify factors for community-based health insurance enrolment in South Achefer District.

Design

Comparative, cross-sectional study.

Settings

Community-based.

Participants

A total of 652 selected households (326 insured and 326 uninsured households) participated in the study.

Methods

A two-sample t-test (for proportions) and χ² (for categorical data) were computed.

Main outcome measure

Utilisation of healthcare.

Results

There was a significant difference in the rate of healthcare utilisation between insured (50.5%) and uninsured (29.3%) households (χ²=27.864, p<0.001). Significant variations of enrolment status in community-based health insurance were observed in the following variables: educational status, family size, occupation, marital status, travel time to the nearest health institution, perceived quality of care, first choice of place for treatment during illness and expected healthcare cost of a recent treatment.

Conclusions

Utilisation of health services among insured households with community-based health insurance was higher. Educational status, family size, occupation, marital status, travel time to the nearest health institution, perceived quality of care, first choice of place for treatment during illness and expected healthcare cost of a recent treatment should be emphasised to enhance community health insurance enrolment.

Strengths and limitations of this study.

• This is the first comparative, cross-sectional study in Achefer, North-West Ethiopia, which assessed healthcare utilisation among insured and uninsured rural households.

• The study identified factors linked with enrolment in a community-based health insurance that will be pertinent in expanding community-based health insurance scheme enrolment.

• As this study was cross-sectional, the factors do not establish temporal relationship; therefore, inference of causation is not possible.

Background

In the lower socioeconomic group of the society, out-of-pocket medical expenditure results in massive financial barriers and impoverished life in the households. Globally, approximately 44 million households (over 150 million people) face financial difficulties due to healthcare expenditure. Consequently, about 25 million households are in deep poverty each year.^{1 2} Over 90% of healthcare financial difficulties and their consequences occurred in Sub-Saharan African countries, where resources are limited.^{3 4} Seven to thirteen per cent of households in six Middle East and North African countries suffered from catastrophic medical expenditure.⁵

The economic burden of direct payments worsens among people in the lower income groups, who struggle to cover daily consumption of food and shelter expenses. In African countries, because of limited health insurance coverage, healthcare demands out-of-pocket payments by borrowing and selling properties.⁶ The level of direct healthcare spending accounted for 42% in Kenya,⁷ 27% in hospitals of Ghana^{8 9} and 37% in Ethiopia.¹⁰

Countries should reduce direct payments of healthcare by introducing health insurance scheme.¹¹ Studies identified that health insurance increases healthcare utilisation.^12–14 In Ethiopia, inpatient healthcare utilisation was 6%, which is one of the lowest health service utilisation rates in Sub-Saharan African countries.⁶ For the last several years, unimproved healthcare services and financial burdens of healthcare are the main issues of Ethiopian people.¹⁵ Only about 1.2% of the citizens had health insurance from both private and public agencies.¹⁶

Generally, health insurance coverage in Sub-Saharan countries is low.¹⁷ The existing health insurances covered almost exclusively the formal sectors, which accounted for 10% of the population.¹⁸ The majority of poor informal sector workers and rural self-employed residents in Africa had never accessed social protection related to health insurance programmes.¹⁹

Community-based health insurance (CBHI) scheme is an emerging alternative to increase primary healthcare access.²⁰ There is an increasing interest in the role of the CBHI schemes in improving equity and access to essential healthcare of the poor, particularly informal sector workers.^{7 10}

Ethiopia has been implementing the CBHI scheme since 2011 to promote health of poor rural residents. However, all rural households are not enrolled in CBHI.^{4 10 16} The meaning of enrolment in CBHI and the link between CBHI with healthcare utilisation are not well described. Hence, this study aimed to identify differences in enrolment in CBHI and to describe the link between CBHI enrolment and healthcare utilisation in the rural communities of Achefere District.

Methods

Study design and setting

A community-based, comparative, cross-sectional study was conducted from February to March 2016 in a rural community of South Achefer District. The district is located in West Gojjam Zone Administration, North-West, Ethiopia. South Achefer District is divided into 20 kebele administrations. In the district a total of 36 204 households were located, of which 12 612 (34.8%) households were enrolled in CBHI. Participants who were insured in the CBHI scheme did not pay direct healthcare costs. However, the healthcare cost is calculated for the purpose of healthcare financing; then the CBHI agency reimburses the healthcare cost to governmental health institutions where patients are receiving healthcare.¹⁰ In the setting of this study, subsistence agriculture is the main economic activity, and the income of majority of the households depends on agriculture.²¹

Inclusion criteria

Household heads older than 18 years old were included in the study.

Exclusion criteria

Either household head or spouse who was unable to communicate during data collection was excluded from the study.

Sampling and sample size determination

The sample was estimated using two population proportion formula with the following assumptions: 80% statistical power with a level of significance at 5%, insured to uninsured ratio of 1:1, and the proportion of health service utilisation was 35% for the insured households and 20% for the uninsured household.²² The calculated sample sizes were 296 from insured households and 296 from uninsured households. Tolerable non-response rate was 10%; hence, the sampling procedure was a multistage sampling: the first stage was kebeles and then the second stage was households within the kebeles; we used a design effect of 2. The resulting sample size was 652 households (326 enrolled households and 326 unenrolled households).

Sampling procedure

Multistage sampling procedure was employed. From South Achefer District, six kebeles were randomly selected. A total of 652 households were included in the study; sampled households were proportionally allocated to each sampled kebele. Furthermore, in each kebele households were proportionally allocated based on insurance status (insured and uninsured) and were systematically selected.

Data collection procedure and studied variables

Ten diploma graduate data collectors were recruited and trained. Data collectors were supervised by two trained nurses. Face-to-face interviews with the household heads were conducted using structured questionnaires that were prepared in English, translated into Amharic and then back-translated. Pretesting was conducted for consistency and ease of understanding. We measured all variables at the household level; enrolment in CBHI is at the household level. The dependent variables were health services utilisation and enrolment in CBHI. The independent variables were sociodemographic factors (age, sex, education, family size); healthcare access-related factors (distance from health institutions, travel time, insurance membership, wealth); and health perception and healthcare needs variables (first choice of place for treatment, sickness of family member and health-seeking behaviour) as adapted from the Ronald Andersen behaviour model.²³

Operational definitions

Utilisation of healthcare was measured as the number of visit/s made by at least one household member at least once in the previous 6 months for health services (diagnostic or treatment). Perceived health status was the respondent’s report about their health status, which was assigned with numerical values according to the following scale: very good=5, good=4, medium=3, poor=2 and very poor=1; then the value was recategorised into good, medium and poor. Perceived quality of care was the extent of the respondent’s view on the quality of healthcare delivery; it was measured by Likert scale questions: very low, low, neutral, high and very high. Chronic illness is a disease condition that lasts more than 3 months. Wealth index was assessed by asking the following components of assets: livestock, crops production, infrastructure (radio, modern bed, mattress, phone, water pump, modern stove), latrine, housing condition (number of room, roof) and total farm size. Household wealth index was computed using principal component analysis. Although there were large data sets, principal component analysis is a technique to reduce the dimensionality of large data sets. Wealth index was categorised as poor, medium and rich. The wealth index of study households ranges from poor to rich. Households indexed in the muster book of the CBHI schemes were recruited as insured, while households which were not indexed to the CBHI schemes were recruited as uninsured. Muster book is a registration book that indicates whether a household is a member of a CBHI or not.

Data analysis and management

Data were entered into Epi Info and transported to STATA version 9 for analysis. Frequencies and proportions were used to describe the study households in relation to the studied variables. Data were presented using tables. To compare households’ characteristics between insured and uninsured households, t-value with 95% CI was computed using two-sample (independent sample) t-tests. χ² test was computed to compare healthcare utilisation between insured and uninsured households.

Ethical considerations

Permission to conduct the study was obtained from each kebele administration involved in the study. The data collectors read the informed consent to study participants who were not able to read and write. Study participants who were able to read were offered to read the informed consent sheet. Then written informed consents were obtained from all study participants either by finger printing or signing on the informed consent sheets.

Patient and public involvement

No patients or the public were directly involved in the development of the research questions and outcome. No patients were involved in recruitment and design of the study. However, the study participants and administrative officials are informed about the research questions and study objectives. The findings of this research have been planned to be disseminated to Amhara Regional Health Bureau, Ethiopia; furthermore, the results of this research will be disseminated to several stakeholders after being published in scientific journals.

Results

A total of 594 households participated in the study; the response rate was 91.1%, and a subtotal of 297 households were insured.

Sociodemographic characteristics

Majority of the respondents were male, both among the insured (69.02%) and the uninsured (62.96%). Educational status, family size, occupation and marital status showed significant difference between the insured versus the uninsured households, but the variables age and sex did not show variation between the insured and uninsured households (table 1).

Table 1.

Sociodemographic characteristics of insured and uninsured study participants in South Achefer District, March 2016 (n=594)

Variables	Insurance status of households		Difference (%)	P values
	Insured n (%)	Uninsured n (%)
Age (years)
18–24	8 (2.6)	6 (2)	0.69	0.794
25–44	126 (42.42)	113 (38)	4.42	0.277
45–65	135 (45.45)	139 (46.8)	−1.35	0.742
>65	28 (9.43)	39 (13.1)	−3.67	0.154
Sex
Male	205 (69.02)	187 (62.96)	6.06	0.119
Female	92 (30.98)	110 (37.04)		0.991
Education
Unable to read and write	175 (58.92)	204 (68.7)	−9.78	0.013*
Able to read and write	50 (16.84)	33 (11.1)	5.74	0.034*
Primary education and above	72 (24.24)	60 (20.2)	4.04	0.201
Family size
<5	110 (37.04)	142 (47.8)	−10.76	0.008*
≥5	187 (62.96)	155 (52.2)		0.005*
Occupation
Farmer	6 (2.02)	17 (5.7)	−3.68	0.019*
Merchant	290 (97.64)	272 (91.6)	6.04	0.01*
Others†	1 (0.34)	8 (2.69)	−2.35	0.019*
Marital status
Married	265 (89.22)	226 (76.1)	13.12	<0.001*
Single	33 (11.11)	70 (23.57)

*Statistically significant at p<0.05.

†Others: housewife, unemployed.

Healthcare access-related factors

Ninety one per cent of the insured and 87.2% of the uninsured households were receiving health services from health centres. Travel time to the nearest health institution varies significantly with a household’s CBHI enrolment status. However, the type of nearby health institution and household wealth index did not vary significantly with a household’s CBHI enrolment status (table 2).

Table 2.

Healthcare access-related factors and enrolment of study participants in community-based health insurance in South Achefer District, March 2016 (n=594)

Variables	Insured n (%)	Uninsured n (%)	Difference %	P values
Nearest health institution
Health centre	271 (91.25)	259 (87.2)	4.05	0.113
Others†	26 (8.75)	38 (12.8)
Distance from health facility in KM Killo metres
<5	156 (52.53)	165 (55.6)	3.07	0.46
≤5	141 (47.47)	132 (44.4)
Travel time in hours
<1	183 (61.62)	138 (46.8)	14.82	<0.001*
≥1	114 (38.38)	159 (53.2)
Household wealth index
Poor	93 (31.31)	105 (35.4)	−4.09	0.297
Medium	97 (32.66)	101 (34)	−1.34	0.397
Rich	107 (36.03)	91 (30.6)	5.43	0.164

*Significant at p<0.05.

†Others: private clinic, hospital.

Health perception and healthcare needs

Fifty one per cent of the uninsured respondents were identified as not healthy, while 39% of the insured respondents were identified as not healthy. The study variables perceived quality of care, first choice of place for treatment during illness and expected healthcare cost of a recent treatment were significantly different among the insured and uninsured households. However, perceived health status, current health status and chronic illness were not significantly different with respect to enrolment in CBHI by households (table 3).

Table 3.

Perceived needs to community-based health insurance enrolment among respondents in South Achefere, March 2016 (n=594)

Variables	Insured n (%)	Uninsured n (%)	Difference	P values
Perceived health status
Poor	39 (43.3)	51 (56.7)	13.4	0.28
Medium	127 (45.8)	150 (54.2)	8.4	0.06
Good	131 (57.7)	96 (42.3)	15.4	0.05
Perceived quality of care
Very low	20 (6.75)	31 (10.4)	3.65	0.108
Low	110 (37.04)	137 (46.1)	9.06	0.025*
Neutral	21 (7.07)	42 (14.1)	7.03	0.05
High	115 (38.72)	69 (23.2)	−15.52	<0.001*
Very high	31 (10.44)	18 (6.1)	4.34	0.02*
Current health status
Healthy	258 (86.87)	246 (82.83)	4.04	0.83
Not healthy	39 (13.13)	51 (17.17)	−4.04	0.17
First choice of place for treatment during illness
Health institution	279 (93.94)	221 (74.41)	19.53	<0.001*
Traditional healer	18 (6.06)	76 (25.59)	−19.53
Chronic illness
Yes	59 (19.9)	45 (15.2)	4.72	0.131
No	238 (80.1	252 (84.8)	−4.7	0.869
Expected healthcare cost of a recent treatment
<500 birr	211 (71.04)	292 (98.32)	−27.28	<0.001*
≥500 birr	86 (28.96)	5 (1.68)

*Significant at p<0.05.

Utilisation of healthcare service and CBHI

In the past 6 months prior to data collection, households who were enrolled in CBHI (50.5%) were more likely to use healthcare services than households who were not enrolled (29.3%) (χ²=27.86, p<0.001) (table 4).

Table 4.

Healthcare utilisation and community health insurance enrolment in South Achefer District, March 2016 (n=594)

Variables	Health service utilisation		χ2	P values
Insurance status	Yes n (%)	No n (%)
Non-insured	87 (29.3)	210 (70.7)	27.864	<0.001*
Insured	150 (50.5)	147 (49.5)

*Statistically significant at p<0.05.

Discussion

This study aimed to compare differences in healthcare utilisation between CBHI member households and non-member households, and the study identified factors for CBHI enrolment. Consistent with previous studies,^{8 24 25} in this study the households enrolled in CBHI were more likely to use health services than the households not enrolled in CBHI. The health service utilisation rate was higher among households who were members of a risk-sharing institution (tax-based health insurance, national health insurance, social health insurance and private health insurance).^{17 26–28} Even if the patient wants to visit a healthcare institution for diagnostic and therapeutic purpose, direct payment for healthcare hinders utilisation of health services.^29–32

A research finding identified that a higher number of study subjects with chronic illnesses were observed among the insured households than the uninsured households⁸; this finding is similar to our research finding. Adverse selection is a critical concern for voluntary-based health insurance. CBHI is a targeted subsidy to the poor households. At the same time this targeted subsidy to the highest risk group increases adverse selection. Therefore, the plan to bridge the financial gap due to adverse selection is crucial in continuing CBHI service to the community.³³

On the other hand, this study showed that the proportion of healthy study subjects during time of data collection was higher among the insured respondents than the uninsured respondents, which was consistent with a study conducted in Nigeria.³⁴ In our study settings, household heads were categorised as healthy or not healthy using perceived health status assessment. Often, household members who were relatively healthier, that is, household members who did not develop a chronic disease, became active household heads and underwent the study interview. The proportion of perceived health status among the interviewed household heads could be higher even though non-household head members with chronic disease were living in the households. Similar to our research finding, in Ghana respondents who perceived their own health condition as good were more likely to be enrolled in the national health insurance scheme.³⁵ A higher proportion of interviewed household heads who were enrolled in a health insurance could have better awareness and practice in disease prevention, health-seeking behaviour and general health; consequently, cumulative perceived health status would be better among enrolled respondents.

Our study showed that a higher proportion of household heads whose wealth index category was the richest were enrolled in CBHI compared with households whose wealth index category was the poorest.³⁶ However, the difference was not statistically significant. In Ethiopia the CBHI scheme payment per year to enrol in CBHI is less than $10.³⁷ The poorest can afford the year-based payment to enrol in CBHI. Therefore, in this circumstance wealth could not be a factor for enrolment in CBHI. In descriptive statistics a higher proportion of the wealthiest households were enrolled in CBHI than the households whose wealth indexes were categorised as poor. This could mean better awareness about the merit of CBHI among the higher wealth index group than the lower wealth index group. This study has limitations. As this study was cross-sectional, the factors do not establish temporal relationship. Therefore, inference of causation is not possible. This study did not use matching criteria for insured and uninsured households.

Conclusion

Utilisation of health services among insured households in CBHI was higher. Educational status, family size, occupation, marital status, travel time to the nearest health institution, perceived quality of care, first choice of place for treatment during illness and expected healthcare cost of a recent treatment should be emphasised to enhance community health insurance enrolment, which leads to universal health coverage.