Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Health care seeking behaviour and utilisation in a multiple health insurance system: does insurance affiliation matter?

Health care seeking behaviour and utilisation in a multiple health insurance system: does... Background: Many countries striving to achieve universal health insurance coverage have done so by means of multiple health insurance funds covering different population groups. However, existence of multiple health insurance funds may also cause variation in access to health care, due to the differential revenue raising capacities and benefit packages offered by the various funds resulting in inequity and inefficiency within the health system. This paper examines how the existence of multiple health insurance funds affects health care seeking behaviour and utilisation among members of the Community Health Fund, the National Health Insurance Fund and non-members in two districts in Tanzania. Methods: Using household survey data collected in 2011 with a sample of 3290 individuals, the study uses a multinomial logit model to examine the influence of predisposing, enabling and need characteristics on the probability of seeking care and choice of provider. Results: Generally, health insurance is found to increase the probability of seeking care and reduce delays. However, the probability, timing of seeking care and choice of provider varies across the CHF and NHIF members. Conclusions: Reducing fragmentation is necessary to provide opportunities for redistribution and to promote equity in utilisation of health services. Improvement in the delivery of services is crucial for achievement of improved access and financial protection and for increased enrolment into the CHF, which is essential for broadening redistribution and cross-subsidisation to promote equity. Keywords: Health insurance, Health care utilisation, Health seeking behaviour, Equity Background which to achieve universal coverage given the con- Health insurance has emerged as a key instrument in straints of enforcing universal mandatory coverage [2]. current health financing reforms of middle and low in- However, when the risk pools are fragmented, this also come countries aimed at achieving universal coverage. causes variation in the potential access, health care seek- The health insurance systems of these countries are ing behaviour and utilisation of health services. This is often characterised by multiple health insurance funds likely because apart from reducing the financial barriers covering different population groups. When mechanisms associated with the cost of health services, health insur- to promote cross-subsidies across funds exist within the ance also influences health care seeking behaviour health insurance system, the risk pools are referred to as (whether, when, from where care is sought for an illness) integrated. Without such mechanisms the risk pools are by preventing delays, self-treatment and use of alterna- referred to as fragmented [1]. Arguably, using multiple tive forms of care [3]. In addition, structural features of health insurance funds is the most practical means with the health insurance system such as contribution levels, eligibility and benefit entitlements determine who is cov- * Correspondence: e_chomi@yahoo.com ered as well as the quality, type and quantity of services Equal contributors covered. Hence the way health insurance system is orga- Muhimbili University of Health and Allied Sciences, Dar-es-Salaam, Tanzania Aarhus University, Aarhus, Denmark nised is likely to influence health care seeking behaviour Full list of author information is available at the end of the article © 2014 Chomi et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 2 of 11 http://www.equityhealthj.com/content/13/1/25 and utilisation of health services. Also, the differential Against this background, this paper intends to exam- revenue raising capacities and benefit packages offered ine the effects of fragmented risk pooling on health care by the various insurance funds in a fragmented system seeking behaviour and utilisation of CHF and NHIF are likely to result in varying degrees of access, health members and non-members in two districts in Tanzania. care seeking behaviour and utilisation of health services Specifically we aim to examine the differences in health [1,4]. Furthermore, fragmentation results in inefficiently care seeking behaviour and utilisation between CHF, high administrative costs, which may have an impact on NHIF and non-members. Bivariate and multivariate ana- the ability of the health system to achieve its policy ob- lyses are used to study the relationship between mem- jectives of financial protection [5,6]. bership status and the decision and timing to seek care, Existence of fragmentation also results in a tiered health and choice of provider given a set of predisposing, enab- system, which is inequitable [2]. The experience of many ling and need characteristics. The remainder of this sec- Latin American countries exemplifies this, where social tion provides a discussion of a framework for health care health insurance for formal workers co-existed with na- utilisation. Section two describes the methods and the tional health services delivered directly through ministries data used in this study and section three presents the re- of health to provide the poor and informal workers with sults. Section four discusses the results and the final sec- health service coverage. This led to a tiered system whereby tion provides some conclusions of the study in terms of formal sector workers enjoyed access to a wide range of policy and further analysis. high quality services while the rest had access to a less gen- erous benefit package while incurring higher costs due to Health care seeking behaviour and utilisation co-payments or excluded services [7-10]. Thus, health in- One of the most frequently used frameworks for the surance systems with fragmented risk pools lack the neces- analysis of health care utilisation is Andersen’s behav- sary conditions for cross-subsidisation to promote financial ioural model of health care use [15-21]. This framework protection and equitable utilisation of health services [11]. assumes that utilisation of health care is influenced by Tanzania has two predominant health insurance funds, the predisposition, the ability and the need to use health the Community Health Fund (CHF) and the National services [22,23]. Health Insurance Fund (NHIF). The NHIF is mandatory Predisposing factors relate to the propensity to utilise for public sector employees covering 7.2% of the population health services and include individual characteristics that while the CHF is voluntary and district based for the rural are not directly related to health care utilisation but population with coverage of about 6.6% [12]. Other insur- rather influence the likelihood of utilisation. These char- ance funds include an urban equivalent of CHF for the in- acteristics can be categorised as: demographic, social formal population, ‘Tiba Kwa Kadi’ (TIKA) and the Social structure and health beliefs [24]. Demographic charac- Health Insurance Benefit (SHIB) for members of the Na- teristics include age and sex, which represent biological tional Social Security Fund (NSSF). There are also various factors that affect the likelihood that an individual will private health insurance funds mostly covering those in the need health services. Social structure represents the fac- formal sector through their employers and micro-insurance tors that determine the status of an individual in the so- schemes which cover mostly informal sector workers ciety as well as the physical and social environment. The [13,14]. most common measures of social structure are educa- The health insurance system is fragmented, with no tion level, occupation and ethnicity. Health beliefs are transfers between the risk pools despite the differential the attitudes, values and knowledge that an individual health care needs and revenue bases. NHIF members are may have about health and health services that may in- entitled to a relatively comprehensive package of health ser- fluence utilisation of health services [23]. vices which include specialised services that can be Enabling characteristics describe the means that indi- accessed from government and accredited private primary, viduals have at their disposal with which to utilise health secondary and tertiary care providers. In contrast CHF services. This is based on the argument that even though members are entitled to a package of health care services an individual may be predisposed to utilise health ser- which they can access from primary care providers [13]. vices, certain factors must be in place to enable actual This implies that members of CHF and NHIF have varying use. These include income, health insurance status and degrees of access to health care hence it is likely that this availability of health services. Usually residence (urban/ may influence their health care seeking behaviour and util- rural) and distance are used as proxy measures for avail- isation. While some degree of cross-subsidisation between ability of health services. Need characteristics are the the healthy and the sick occurs within the CHF and NHIF, direct determinants of health care use which include self the lack of cross-subsidies across the funds limits the extent reported and evaluated morbidity [22]. to which resources can be redistributed to promote equit- Health insurance is the primary variable of interest in able utilisation. this study, due to the key role in improving access to Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 3 of 11 http://www.equityhealthj.com/content/13/1/25 health care by reducing financial barriers to utilisation of This approach was used since there are few NHIF mem- health services. It is therefore expected that health insur- bers. Non-member households were randomly selected ance should positively influence the probability of util- from the village household register in each of the se- isation. The fragmentation of the health insurance lected villages. All CHF and NHIF households were system and the differential benefit packages between the omitted from the village register using the list obtained CHF and NHIF implies that members of these two funds from the facility and District Council respectively before will have differential access to health care, which will in- selection of non-member households. In each household herently influence their choice of provider. From this all members were interviewed. perspective, we expect that choice of provider will be in- Data used for analysis was collected as part of a larger fluenced by insurance affiliation. Predisposing and need study, hence estimated sample size calculations were variables are used as control variables. based on all the key study variables and the maximum sample size estimate was used, since this would be suffi- Methodology cient for the analysis of all key variables [27]. Hence, Study setting, design and data collection while we obtained estimates based on the key variables Data for this study was obtained from Kongwa and for this analysis, they were not sufficient for the analysis Mpwapwa districts in Tanzania over a period of eight of other key variables. The sample size calculation was weeks between July and September 2011. The two dis- based on the assumptions that there would be 80% tricts were selected due to their different levels of CHF power to detect a 25% difference between CHF and enrollment, and for convenience in terms of logistics NHIF households in the number of facility visits per year. and costs. Kongwa has a total of 63,612 households We used the proportion from a similar study, which of which 5,800 (9%) are registered with CHF [25]. reported about 50% of insured households having at least Mpwapwa has a total of 78,812 households of which one facility visit per year [28] and assumed the propor- 15,540 (18%) are registered with CHF [25]. The prime tion of NHIF households to be higher. This resulted in a economic activity in both districts is agriculture and live- sample of 729 households (243 per group). Estimating a stock keeping. non-response of 5%, final sample size was adjusted to 766 households. Using the average household size of 5 Sampling method and sample size calculation persons reported by the 2010 TDHS, this represents a For the purposes of this study a household is defined as a sample size of approximately 3830 individuals. person or group of people related or unrelated who live to- A pre-tested structured questionnaire was adminis- gether and share a common pot of food and who share the tered to the household head or spouse. Data was col- same membership card (for CHF) or are dependents of the lected on demographic characteristics, employment, same principal member (NHIF households). This was education level, family size, membership status, house- adapted from the 2010 Tanzania Demographic and Health hold ownership of assets and consumer durables, pres- Survey (TDHS) definition of a household [26]. The study ence of chronic and acute illnesses, general health population comprised of all individuals living in the house- status, health care seeking behaviour and utilisation of holds in the two districts which met this definition. In each health services. Three return visits were made to house- district a multi-stage sampling approach was used to select holds where members were not available for interview first wards, then villages followed by hamlets and eventu- during the first visit, resulting in a response rate of 85%, ally households. Due to difficulties in identification of with a sample size of 3290 individuals from 695 households by membership status from the village house- households. hold register , equal numbers of households were selected from listings of each membership category as follows: CHF Study variables households were randomly selected from the CHF register For the bivariate analysis, the decision and timing of seek- book kept in the health facilities in each ward. This was be- ing care was compared across the CHF, NHIF and non- cause health facilities are registration points for CHF regis- members. The decision to seek care relates to whether or tration. The health facilities were selected based on not care was sought for an illness experience during the whether the facility catchment area falls within the selected four weeks recall period. Timing relates to the time elapsed hamlets. The selection was made from members registered between the onset of symptoms of illness and seeking care from September 2010 to September 2011, to ensure only (same day, less than 1 week, more than 1 week). current CHF members were included. For the multivariate analysis, choice of provider, de- For NHIF households, a list of all Government institu- fined as the place of first contact following an illness tions in the selected wards or villages was obtained from during the four weeks recall period (public hospital, pri- the District Council, from which all available (at the time vate health facility, public health centre/dispensary or of the study) NHIF principal members were selected. traditional healer/self medication) was the dependent Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 4 of 11 http://www.equityhealthj.com/content/13/1/25 variable. The alternative of traditional healer or self The MNL model assumes that the odds of choosing medication refers to those who sought treatment outside between two alternative choices do not depend on which the home from a traditional healer, drugstore or phar- other choices are available (the Independence of Irrele- macy. This differs from individuals who did not seek vant Alternatives (IIA) assumption) [33]. We employed care but instead used home remedies or those who de- Hausman-McFadden (HM) and Small-Hsiao (SH) tests layed seeking care and opted to start with home remed- to validate this assumption. Both tests returned non- ies first. Traditional healer and self medication was later significant results (HM-ρ=0.112; SH-ρ=0.112), indicating merged since both choices represent alternative sources that the model is appropriate. The use of alternative of care. In addition the choice of traditional healer models that relax the IIA assumption such as the multi- accounted for only 1% of those seeking care. nomial probit, nested logit and mixed logit models is Drawing on Anderson’s 1995 model [23], independent limited by their computational difficulties and for the variables used in this study include predisposing, enabling multinomial probit, the need for a particular data struc- and need characteristics of individuals. Predisposing charac- ture [32]. teristics included age (0–5, 6–14, 15–49, 50–59, 60+ years), It is also possible that the same factors that influence sex (male, female) and education level of household head health care utilisation could also influence the purchase (no education, primary education, secondary education, of health insurance, implying that the health insurance above secondary education). Perceived adequacy of staff variable is endogenous [33-35]. The mandatory nature of and perceived availability of drugs (yes, no) were included NHIF and household basis of membership for the CHF as proxies for attitudes towards health services as expressed reduce the effect of selection bias in our study. However, from general questions on health status and utilisation. possible endogeneity of the health insurance variable Enabling characteristics included were household char- was tested using the Durbin-Wu-Hausman test, which acteristics that were assigned to an individual according uses an instrumental variable to test whether the predic- to the household to which he/she belonged. These were tors are correlated with the error term. A non- membership status (CHF, NHIF, non-members), resi- significant test result indicates that none of the predictor dence (urban, rural), distance to facility (less than or variables are endogenous [36]. We used relationship to more than 5 km) and wealth status (lowest to highest the head of household as the instrumental variable and wealth quintile). Owing to the complexities of determin- the test was not significant (ρ=0.315) implying that the ing actual income, Principal Components Analysis health insurance variable was exogenous. Choice of the (PCA) was used to develop an asset index that grouped instrumental variable was based on its influence on households into quintiles based on ownership of assets health insurance membership but not on health care and durable goods [29,30]. The asset index of the house- utilisation, which is a criterion for instrumental variable hold was used to represent socio-economic status. For selection. Being related to the household head makes the need characteristics we used self reported illness, de- one eligible for insurance membership, but does not in- fined as the experience of illness or injury lasting for a fluence whether or where care will be sought for an month or less (acute) or experience of an illness lasting illness. for three months or more (chronic). Since individuals in the sample were obtained from households, the observations of each individual are not Analysis independent of each other, resulting in an under- Bivariate and multivariate analysis was used to study the re- estimation of standard errors and making significance lationship of predisposing, enabling and need characteris- tests used in the analysis invalid. The effect of clustering tics and health care seeking behaviour and utilisation. Chi has been accounted for in the analysis, using clustered square tests were used to study the relationship between robust standard errors which increase the variability be- membership status and health care seeking behaviour along tween individuals within cluster [37-39]. two dimensions: the decision to seek care for an illness and timing. Multinomial logistic (MNL) regression was used to Ethical considerations estimate the choice of provider given a set of predisposing, Ethical approval was sought from the Research and Ethics enabling and need characteristics. This model was selected Committee of Muhimbili University of Health and Allied based on the nature of the dependent variable and the abil- Sciences. Following ethical approval, permission to con- ity of the model to estimate all choices in a single equation duct data collection was obtained from the Regional [31,32]. Since the aim was to find out whether insured indi- Administrative Secretary (RAS) of Dodoma and District viduals will choose a provider where they can use their in- Administrative Secretary (DAS) of Kongwa and Mpwapwa. surance card, the health centre/dispensary category was Respondents were informed of the research objectives and used as a reference (both CHF and NHIF members can use were asked to participate in the study. Those who agreed their cards at this level). were asked to sign a consent form. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 5 of 11 http://www.equityhealthj.com/content/13/1/25 Results education, while the majority of CHF (72%) and non- Descriptive characteristics of the sample members (64%) attained primary education (ρ<0.05). Table 1 presents the descriptive characteristics of the The majority of NHIF households are relatively wealthy study sample. About 28% of the respondents were from (39% highest, 40% fourth wealth quintile) compared to NHIF households, 38% from CHF households and 34% CHF (29% lowest, 31% second wealth quintile) and non- from non-member households. More than 50% of NHIF member households (33% lowest, 27% second wealth members were in the 15–49 year age group, with less quintile, ρ<0.05). than 10% aged 0–5 years and less than 3% aged 60 years and above. In contrast 43% each of CHF and non- Health care seeking behaviour and utilisation members were aged 15–49 years, 15% aged 0–5 years Table 2 illustrates health care seeking behaviour by and 4% aged 60 and above (ρ<0.05). A higher proportion membership status. During the recall period of four of CHF households (44%) had more than five members, weeks prior to the survey, 30 percent reported having compared to NHIF (28%) and non-member households had at least one spell of illness. Of those individuals, 75 (31%, ρ<0.05). percent reported having sought care for their illness. NHIF head of households were more educated, with 63% having attained secondary or above secondary Delays in seeking care The majority of CHF (57%) and NHIF (66%) members were more likely to seek care on the same day they fell Table 1 Individual characteristics by membership status, ill, while more than 40% of non-members were more Kongwa and Mpwapwa 2011 (%) likely to experience delays in seeking care (ρ<0.05 ). Variable NHIF CHF Non-members The reasons for delaying to seek care and/or not seek- N = 931 N = 1242 1117 ing care are presented in Table 3. Among those who de- Age*** layed seeking care, the most common reason reported 0-5 9.1 14.5 14.8 was “wait and see if illness progresses further” for NHIF 6-14 25.2 33.4 31.3 (40%) and CHF (44%) members. Among non-members 15-49 55.2 42.9 42.9 lack of money to pay for treatment was the main reason, reported by 34%. Reasons for not seeking care differed 50-59 8.7 4.9 6.5 significantly by membership status. The main reasons 60+ 1.8 4.3 4.4 were availability of own medicine/home remedies, re- Sex ported by 43% of NHIF and 38% of CHF members and Male 47.1 48.6 46.9 lack of money to pay for treatment, reported by 51% of Female 52.9 51.5 53.1 non-members (ρ<0.05). Education (head)*** Choice of Provider No education 1.4 20.7 30.7 The most common choice of provider was the public Up to Primary 8.1 71.9 63.9 health centre/dispensary; 43% of all individuals seeking Up to–Secondary 27.9 6.5 5.0 care chose this source, followed by traditional healer/self Above secondary 62.6 0.7 0.4 medication (28%), public hospital (18%) and private Household size*** Mean = 4.7 4.0 5.2 4.7 Table 2 Health seeking behaviour by membership status, 1-5 members 72.3 56.2 69.3 conditioned on reporting illness (%) >5 members 27.7 43.8 30.7 Variable NHIF CHF Non-members Wealth Status*** N = 931 N = 1242 N = 1117 Lowest 2.5 29.4 33.2 Reported illness (yes) 31.3 30.5 28.6 Second 3.7 30.8 27.2 Sought care for illness?*** N = 291 N = 379 N = 320 Third 14.7 22.7 22.7 Yes 81.1 77.3 65.6 Fourth 40.2 12.7 10.3 No 18.9 22.7 34.4 Highest 38.9 4.4 6.5 Timing of care** N = 236 N = 293 N = 201 Distance to facility Same day 66.1 57.3 52.7 1-5 km 91.9 95.4 95.4 <1 week 31.4 37.8 40.8 >5 km 4.6 8.1 4.7 >1 week 2.5 4.8 6.5 ***ρ<0.01; based on Chi square test. **ρ<0.05; ***ρ<0.01; based on Chi square test. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 6 of 11 http://www.equityhealthj.com/content/13/1/25 Table 3 Reasons for delays and/or not seeking care by dispensaries, compared to 28% of NHIF members and membership status, conditioned on reporting illness (%) 35% non-members (ρ<0.05). Private facilities were pre- Variable NHIF CHF Non-members ferred by 22% of NHIF members, 5% of CHF members and 10% of non-members. Reasons for delays*** N = 77 N = 120 N = 90 No money 9.1 9.2 40.0 Reasons for choice of provider Distance 6.5 1.7 1.1 Reasons for choice of provider are categorised into those Self treatment (home) 18.2 16.7 12.2 for choice of formal care and those for choice of alterna- “Wait and see” 49.4 49.2 33.3 tive forms of care (Table 4). Good quality (availability of Facility closed 10.4 11.7 10.0 drugs, laboratory tests, staff and most likely to find a doctor) was the main reason for seeking care from a dis- Other 11.7 11.7 3.3 trict hospital, reported by 41% of non-members and Reasons for not seeking care*** N = 51 N = 68 N = 97 NHIF members who used this provider. In contrast, 52% Poor quality of services 11.8 1.5 0.0 of CHF members who sought care from a public hospital No money 5.9 10.3 50.5 said it was the only facility nearby (ρ<0.05). Quality rea- No insurance card 1.9 0.0 5.2 sons were also the most important for those who sought Distance 0.0 1.5 0.0 care from a private facility, reported by 73% of NHIF members, 50% of CHF members and 65% of non- Illness not serious 23.5 30.9 11.3 members (ρ<0.05). No one to leave at home/farm 3.9 10.3 1.0 The main reason for seeking care from public pri- Had my own medicine/home 43.1 38.2 26.8 mary care facilities is ‘only facility available nearby’, remedies reported by 61% of NHIF members, 64% of CHF Other 7.4 9.8 5.2 members and 75% of non-members who sought care ***ρ<0.01; based on Chi square test. from this source. Of particular interest is the import- facilities (12%). About 22% of NHIF members who ance of insurance for CHF members, reflected by sought care for an illness went to a public hospital, com- about 30% whose choice of provider was based on pared to 16% of CHF members and 16% of non- the ability to use their insurance card, compared to members. A higher proportion of non-members opted 20% and 18% of NHIF members. for traditional healers/self medication (40%), compared Among those who sought alternative forms of care, to that of NHIF members (29%) and CHF members the main reason reported was that they were more likely (20%, ρ<0.05). Figure 1 illustrates the different choice to receive treatment from this source rather than from patterns by membership status. It shows that 60% of formal sources of care (NHIF 43%, CHF 47%, non- CHF members sought care from public health centre/ members 27%, ρ<0.05). CHF NHIF non member choice of provider public hospital private hospital/HC/D/C public HC/D traditional/ self medication Figure 1 Choice of provider by membership status, conditioned on reporting illness (%). 0 20 40 60 Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 7 of 11 http://www.equityhealthj.com/content/13/1/25 Table 4 Reasons for choice of care by membership status healer/self medication over a public health centre/dispens- and provider, conditioned on seeking care (%) ary. Urban residents are more likely than rural residents to Variable NHIF CHF Non-members choose a public hospital, private facility and self medication N = 171 N = 237 N = 128 over a public health centre/dispensary, regardless of health District hospital (N = 130)*** insurance status. Individuals are more likely to travel long Good quality 40.8 15.2 41.3 distances to seek care from a public hospital rather than a public health centre/dispensary. Only facility available nearby 34.7 52.2 41.2 Insurance card accepted 20.1 30.4 NA The effect of need characteristics Exemption NA NA 8.8 Individuals reporting chronicillness aremorelikelyto Private Health centre/Dispensary (N = 85)*** choose a public hospital or a private facility over a public Good quality 72.6 50.0 65.0 health centre/dispensary. Results indicate that acute illness Only facility available nearby 17.7 42.9 30.0 is not a significant determinant of choice of provider. Insurance card accepted 5.9 0.0 NA Discussion Exemption NA NA 0.0 This paper examined the effects of fragmented risk pool- Public Health center/Dispensary (N = 315)*** ing on health care seeking behaviour and utilisation of Good quality 21.2 5.1 17.8 CHF and NHIF members and non-members. Results Only facility available nearby 60.6 64.2 75.3 suggest that the insured are more likely to seek care and Insurance card accepted 18.2 30.6 NA less likely to experience delays compared to non- members. Lack of money to pay for treatment is a sig- Exemption NA NA 2.7 nificant barrier to seeking care for non-members but not Choice of alternative care (traditional healer/self for CHF and NHIF members, since this was the main rea- medication, N = 209)** son for delays in seeking care or not seeking care at all N = 58 N = 51 N = 70 reported by non-members. This suggests that generally No money 5.2 11.8 21.4 health insurance does improve access to health care by re- Distance 17.2 13.7 8.6 ducing the financial barriers associated with utilisation of More likely to get treatment 43.1 47.1 27.1 health services and is consistent with results reported by Poor quality of services at 8.6 7.8 12.9 Jutting [37], Bronwyn et. al [40], Mensah et al. [41] and formal facility SHIELD [42]. However, members of the NHIF are more No insurance card 0.0 3.9 7.1 likely than CHF members to seek care for an illness and Illness not serious 8.6 3.9 2.7 are also less likely to delay seeking care. This variation cor- roborates findings of Ekman [4] in Jordan who reported a Recognise symptoms, familiar 6.9 3.9 14.3 with drugs higher probability of seeking care among members of the Ministry of health insurance programme compared to Other 8.6 7.8 5.7 other programmes. The variation between CHF and NHIF **ρ<0.05; ***ρ<0.01; based on Chi square test. Percentages may not add up to 100% since only the 3 main reasons members can be explained by the fact that compared to are presented. CHF members, NHIF members are more likely to live near a health facility and have a wider choice of providers Multivariate results compared to CHF members. Table 5 presents results from the multinomial logit Differences were also found in relation to the choice model specification. of provider. Compared to CHF members, NHIF mem- bers are more likely to choose a private facility, trad- The effect of enabling characteristics itional healer or self medication over a public health Persons from wealthier households are more likely to centre or dispensary. Although not significant, results choose a public hospital, private facility and traditional also show that NHIF members are more likely than CHF healer/self medication over a public health centre/dispens- members to choose a public hospital rather than a pub- ary compared to those from relatively poorer households. lic health centre or dispensary. Good quality (availability The positive sign on the coefficient indicates that compared of drugs, laboratory tests, staff and most likely to find a to CHF members, NHIF members are more likely to doctor) was the main reason for seeking care from a dis- choose a private facility or traditional healer/self medication trict hospital and private facility. District hospitals and over a public health centre/dispensary. Likewise, non- private facilities are usually relatively better equipped members are more likely than CHF members to choose a and are more likely to have drugs than dispensaries or public hospital, private health facility and a traditional health centres [25]. Moreover, a non-functional referral Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 8 of 11 http://www.equityhealthj.com/content/13/1/25 Table 5 Multinomial logit estimation results, (public health centre/dispensary as reference) Variable Public hospital Private hospital Traditional/self Coeff. Robust Std Error Coeff. Robust Std Error Coeff. Robust Std Error Age (0–5 ) 6-14 −0.241 0.369 0.113 0.437 0.340 0.302 15-49 0.116 0.340 0.693* 0.413 0.836** 0.304 50-69 0.707 0.548 0.131 0.737 1.010** 0.465 60+ −1.367 1.043 −0.363 1.103 0.648 0.579 Sex (male ) Female 0.023 0.260 0.250 0.299 −0.142 0.200 Education level (no education ) Up to Primary 0.437 0.540 0.378 0.638 0.480 0.405 Up to Secondary −0.224 0.702 −0.964 0.908 −1.766** 0.646 Above Secondary −1.346 0.567 −1.311 0.940 −2.191*** 0.716 Wealth status (lowest ) Second −0.533 0.594 −0.071 0.651 0.239 0.377 Third 0.360 0.490 0.214 0.693 0.811** 0.357 Fourth 1.008 0.626 1.295* 0.728 1.525** 0.553 Highest 1.675** 0.657 1.940** 0.746 1.313** 0.590 Membership status (CHF ) NHIF 0.037 0.679 1.664** 0.770 2.123*** 0.574 Non-member 1.078** 0.411 1.594*** 0.506 1.520*** 0.307 Residence (rural ) Urban 3.721*** 0.453 1.868*** 0.446 1.046*** 0.359 Distance 0.091 0.072 0.075 0.070 0.041 0.062 Self reported morbidity (no illness ) Acute 0.243 0.675 0.773 0.912 0.076 0.548 Chronic 1.695** 0.821 2.201** 1.022 0.980 0.677 Perceived quality of care (poor ) Availability of drugs −0.791** 0.374 −1.085*** 0.356 −0.358* 0.279 Adequacy of staff 0.196 0.371 0.388 0.326 0.386 0.300 Constant −3.931*** 0.894 −4.762*** 1.380 −2.620*** 0.733 *ρ<0.1; **ρ<0.05; ***ρ<0.01; base category. system and no gate-keeping makes it easier for patients ‘insurance card accepted’ (31%), while quality reasons to bypass primary care facilities to seek care from hospi- were reported by only 5% of CHF members. This implies tals [43,44]. Given that the NHIF coverage extends to all that CHF members will choose a provider that is nearby levels of care and to private accredited facilities, it is not and where they can use their insurance card to pay for surprising that members chose providers on the basis of services regardless of quality of care offered. The CHF perceived better quality of care. A study in Indonesia benefit package is limited to primary level care offered at examining the effects of mandatory insurance on access health centres and dispensaries, and CHF registration to carealsofound apositiveeffect of theAskes (for occurs at the health facility, usually one that is nearest public employees) on the use of public providers, while to where the member lives. Hence it is possible that the Jamostek (for private employees) had a positive ef- members link their insurance entitlements to the facility fect on useof bothpublicand privateproviders.This where they registered making this facility the most lo- was attributed to the differential benefit packages by the gical choice. Therefore, unlike NHIF members, for CHF two funds [45]. members, choice of provider involves a trade-off be- The main reasons for choice of provider among CHF tween claiming their entitlements from providers who members were ‘only facility available nearby’ (64%) and may not necessarily provide the desired quality of care Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 9 of 11 http://www.equityhealthj.com/content/13/1/25 and forgoing their entitlements by paying out-of pocket In the Andersen behavioural model of health care util- for perceived superior quality. This may make health in- isation, enabling characteristics such as income and surance an unattractive product, reducing the likelihood health insurance are described as those that are neces- of re-enrolment of CHF members and negatively influ- sary but not sufficient for utilisation [48]. Adequate ence non-members’ decision to join the fund. health infrastructure capable of delivering quality health The variation in health seeking behaviour and choice services has been mentioned as one of the pre-requisites of provider between the CHF and NHIF members is a of successful implementation of health insurance [49-51] reflection of the effects of a fragmented health insurance and achievement of a truly universal health system [52]. system. The lack of a standardised benefit package for The same arguments have been raised by Robyn et. al the members of the two funds implies that while both [3], who found that the effect of health insurance on CHF and NHIF members have better potential access, health seeking behaviour in Burkina Faso was limited, the quality and quantity entitled to respective members owing to poor quality of services. This means that differs. In other words the whether, when, where health achieving basic coverage of the population is meaning- care is sought and the quality and quantity of health ser- less when this coverage does not guarantee access to ser- vices received depends on health insurance affiliation. vices of an adequate quality. Our findings show that the Given the lower revenue raising capacity of CHF com- extent to which health insurance promotes health care pared to the NHIF, it also means that access to health utilisation is dependent on the quality of services offered care is based on ability to pay rather than on need for at formal care providers. When patient expectations are the services. This goes against the principle of equity of not met, it is likely that they will seek alternatives even access and calls for mechanisms that will promote though it means they have to forgo their insurance bene- broader risk sharing and redistribution across the two fits. This has been elucidated by our results, which show schemes and a standardised benefit package. In this way that the majority of CHF and NHIF members who for a given need, insured individuals will enjoy the same sought care from private facilities and from alternative degree of access to health services regardless of insur- sources (self-medication/traditional healer) did so due to ance affiliation or ability to pay. perceived inferior quality in public facilities. Given the One of the goals of expanding health insurance in the voluntary nature of CHF, and its potential for covering Tanzanian health system was to eventually achieve uni- the majority of the population, quality improvements in versal coverage and universal access to health care. For a health services are important to encourage enrolment. system to be truly universal it has to be equitable, grant- Scaling up enrolment into CHF is important for the re- ing access to the same range of services for all based on distributive potential of the scheme and ultimately of an need while requiring payment for these services based integrated health insurance system. on their income [46,47]. This can only be achieved in a health insurance system that is redistributive, such that Conclusion there is risk sharing and cross-subsidisation across insur- This paper examined the effects of fragmented risk pool- ance schemes. Expanding health insurance to cover all ing on health care seeking behaviour and utilisation of population groups without redistributive mechanisms may CHF and NHIF members and non-members. Specific achieve universal coverage but also create a system that areas that were examined were the decision to seek care does not support equity. Our results have shown the in- when ill and the timing and choice of health service pro- equalities in access between the CHF and NHIF members. viders. The findings of this study provide lessons for pol- The expansion of health insurance is a reflection of icy makers in low- and middle-income countries where commitment by the Tanzanian government to achieve multiple health insurance funds have been established to universal coverage. What remains is the development of achieve universal coverage. In particular, addressing the policy framework and design issues that will promote re- challenges of limited risk sharing and cross-subsidisation distribution and cross-subsidisation across the schemes in across multiple health insurance funds remains crucial order to create a health insurance system that is universal for equitable access. These results confirm the import- and equitable. Creating a standardised benefit package is a ance of reducing fragmentation in risk pooling arrange- crucial step in promoting equitable access to health ser- ments, creating the opportunity for risk and income vices. The requirement for a standardised benefit package cross-subsidisation that will also promote the develop- lies in the link between the package, risk structure and ex- ment of a standard benefit package. pected health expenditures of a scheme. Without a stan- Perceived poor quality of health limits the degree to dardised benefit package, redistribution across schemes which the objectives of improved access and financial will reward those with more comprehensive packages at protection can be achieved. Furthermore, poor quality of the expense of schemes with fewer benefits, perpetuating health services serves as a deterrent for enrolment into rather than reducing inequity. voluntary funds, which often represent crucial elements Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 10 of 11 http://www.equityhealthj.com/content/13/1/25 for broadening redistribution and cross-subsidisation to insurance on reducing informal self-care in Burkina Faso. Health Policy Plann 2011, 27(2):156–165. promote equity. 4. Ekman B: The impact of health insurance on outpatient utilisation and expenditure: evidence from one middle-income country using national household survey data. Health Res Policy Syst 2007, 5:6–21. Endnotes 5. Kirigia JM, Preker G, Mwikisa C, Diarra-Nama AJ: An overview of health Principal member is the contributing member of the financing patterns and the way forward in the WHO African Region. NHIF, usually the head of household or spouse. East Afr Med J 2006, 83(8):S1–S28. 6. McIntyre D: Learning from experience: health care financing in low- and In Tanzania, districts are organised into divisions, middle-income countries. Geneva: Global Forum for Health Research; 2006. which in turn are divided into wards. Within each ward, 7. Wagstaff A: Social health insurance re-examined. World Bank Policy Research there are a number of villages, which are also divided Working Paper 4111. Washington DC: World Bank; 2007. into hamlets. Depending on the ward and health infra- 8. Pinto D, Hsiao WC: Colombia: Social health insurance with managed competition to improve health care delivery.In Health systems in low- and structure, one health facility may have a catchment area middle-income countries. Edited by Hsiao WC, Shaw PR. Washington DC: of one or more villages. World Bank; 2007:105–132. Each village has a list of all households registered at 9. Baeza CC, Packard TG: Beyond survival: protecting households from health shocks in Latin America. California: Stanford University Press: Washington DC: the office of the Village Executive Officer. This list is World Bank; 2006. broken down by hamlet but does not show membership 10. Jack W: Health insurance reform in four Latin American countries: theory status. and practice. Policy Research Working Paper, Volume 2492. Washington DC: World Bank; 2000. 11. McIntyre D: What healthcare financing changes are needed to reach Abbreviations universal coverage in South Africa? S Afr Med J 2012, 102(6):489–490. CHF: Community Health Fund; NHIF: National Health Insurance Fund; TIKA: Tiba kwa Kadi (urban equivalent of CHF); SHIB: Social Health Insurance 12. National Health Insurance Fund: Actuarial valuation report of the National Benefit; NSSF: National Social Security Fund; TDHS: Tanzania Demographic Health Insurance Fund. Dar-es-Salaam: M/S iPrint Ltd; 2010. and Health Survey; PCA: Principal Components Analysis; MNL: Multinomial 13. Ministry of Health and Social Welfare: Health financing in Tanzania: where logit; IIA: Independence of Irrelevant Assumptions; RAS: Regional we came from and where we are now. Presented at the National Forum Administrative Secretary; DAS: District Administrative Secretary. for Health Financing. Dar-es-Salaam; 2010. 14. Haazen D: Making health financing work for poor people. Washington DC: World Bank; 2012. Competing interests 15. Young JT, Menken J, Williams J, Khan N, Kuhn R: Who receives healthcare? The Authors declare no competing interests. Age and sex differentials in adult use of healthcare services in rural Bangladesh. World health Popul 2006, 8(2):83–100. Authors’ contributions 16. Brown C, Barner J, Bohman T, Richards K: A multivariate test of an ENC was involved in study design, training of research assistants, review of expanded Andersen health care utilisation model for complementary data collection tools, data collection, data entry, data analysis and drafting and alternative medicine (CAM) use in African Americans. J Altern the manuscript, revising and writing of the final manuscript. PGMM, UE, ADK Complem Med 2009, 15(8):911–919. and KH were involved in study design, review of data collection tools, data 17. Lopez-Cevallos DF, Chi C: Assessing the context of health care utilisation in analysis and revising the manuscript. All the authors have read and approved Ecuador: a spatial and multilevel analysis. BMC Health Serv Res 2010, 10:64. the final version of the manuscript. 18. Jahangir E, Irazola V, Rubinstein A: Need, enabling predisposing and behavioural determinants of access to preventive care in Argentina: Acknowledgements analysis of the national survey of risk factors. PLoS ONE 2012, 7(9):e45053. This research was part of a research program on health insurance in Ghana 19. Glei DA, Goldman N, Rodriguez G: Utilisation of care during pregnancy in and Tanzania, supported by the Danish Consultative Committee on rural Guatemala: does obstetrical need matter? Soc Sci Med 2003, Development Research (Grant DFC 19-08-AAU) within the framework of the 57:2447–2463. Danish development cooperation. The funder was not involved in the 20. Nanjin N, Bennett CM, Luby S: Inequalities in care-seeking for febrile design, data collection, analysis and interpretation of data, or decision to illness of under-five children in urban Dhaka, Bangladesh. J Health Popul publish the results. We also appreciate the Muhimbilii University of Health Nutr 2011, 29(5):523–531. and Allied Sciences, Dar es Salaam for providing ethical approval to conduct 21. Kamgnia B: Use of health services in Cameroon. Int J Appl Econometrics the study. In addition we are thankful for the time spared by the participants Quantitative Studies 2006, 3(2):55–64. who contributed to the successful conduct of this study; the District Councils 22. Andersen R, Newman J: Societal and individual determinants of medical of Kongwa and Mpwapwa and the Ward and Village authorities where the care utilization in the United States. Milbank Q 1973, 51(1):95–124. data was collected who granted permission to conduct the study. 23. Andersen R: Revisiting the behavioural model and access to medical care: does it matter? J Health Soc Behav 1995, 36:1–10. Author details Muhimbili University of Health and Allied Sciences, Dar-es-Salaam, Tanzania. 24. Andersen R: A behavioural model of families’ use of health services.In 2 3 Aarhus University, Aarhus, Denmark. London School of Hygiene and Research Series No. 25. Chicago: Center for Health Administration Studies, Tropical Medicine, London, UK. University of Chicago; 1968. 25. United Republic of Tanzania: Tanzania Health Statistical Abstract. Dar-es- Received: 25 September 2013 Accepted: 12 March 2014 Salaam: Ministry of Health and Social Welfare; 2011. Published: 19 March 2014 26. United Republic of Tanzania: Tanzania demographic and health survey 2010. Dar-es-Salaam: National Bureau of Statistics; 2011. 27. Varkevvisser CM, Pathmanathan I, Brownlee A: Designing and Conducting References Health Systems Research Projects. Volume 1: Proposal Development and 1. Smith P, Witter S: Risk pooling in health care financing: the implications for Fieldwork. Amsterdam: KIT Publishers, Ottawa: International Development health system performance. HNP Discussion Paper. Washington DC: World Research Centre. World Health Organisation-Regional Office for Africa: Bank; 2004. Brazzaville; 2003. 2. McIntyre D, Kutzin J: Revenue collection and pooling arrangements in financing.In Health systems in low- and middle-income countries. Edited by 28. Gnawali DP, Pokhrel S, Sie A, Sanon M, De Allegri M, Souares A, Dong H, Smith RD, Hanson K. New York: Oxford University Press; 2012:77–101. Sauerborn R: The effect of community-based health insurance on the 3. Robyn PJ, Hill A, Liu Y, Souares A, Savadogo G, Sié A, Sauerborn R: utilisation of modern health care services: evidence from Burkina Faso. Econometric analysis to evaluate the effect of community-based health Health Policy 2009, 90(2–3):214–222. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 11 of 11 http://www.equityhealthj.com/content/13/1/25 29. Vyas S, Kumaranayake L: Constructing socio-economic status indices: how to use principal components analysis. Oxford University Press; 2006. 30. Filmer D, Pritchett L: Estimating wealth effects without income or expenditure data-or tears. An application to educational enrolment states in India. World Bank Policy Research Working Paper. Washington DC: World Bank; 2001. 31. Agresti A: An introduction to categorical data analysis. 2nd edition. New Jersey: John Wiley and Sons, Inc.; 2007. 32. Cheng S, Long JS: Testing for IIA in the multinomial logit model. Socio Meth Res 2007, 35(4):583–600. 33. Long JS, Freese J: Regression models for categorical dependent variable using Stata. Texas: Stata Press; 2006. 34. Waters HR: Measuring the impact of health insurance with a correction for selection bias-a case study of Ecuador. Health Econ 1999, 8(5):473–483. 35. Davidson R, MacKinnon JG: Econometric Theory and Method. New York: Oxford University Press, 2004. In Ekman B: The Impact of Health Insurance on Outpatient Utilisation and Expenditure: Evidence from one Middle-income Country using National Household Survey Data. Health Res Policy and Syst 2007, 5:6–21. 36. UCLA: Statistical Consulting Group: Analysing correlated (clustered) data. [http://www.ats.ucla.edu/stat/stata/library/cpsu.htm] 37. Juul S: An introduction to Stata for health researchers.2nd Edition. Texas: Stata Press; 2008. 38. Jowett M, Deolalikar A, Martinsson P: Health insurance and treatment seeking behaviour: evidence from a low-income country. Health Econ 2004, 13:845–857. 39. Jutting JP: Do community-based health insurance schemes improve poor people’s access to health care? Evidence from rural Senegal. World Dev 2004, 32:273–288. 40. Bronwyn H, Goudge J, Ataguba JE, McIntyre D, Nxumalo N, Jikwana S, Chrisich M: Inequities in access to health care in South Africa. J Public Health Policy 2011, 32:S102–S123. 41. Mensah J, Oppong JR, Bobi-Barimah K, Frempong G, Sabi W: An evaluation of the Ghana National Health Insurance Scheme in the context of the health MDGs. Global Development Network Working Paper Series 2010, No. 40. GDN. 42. SHIELD: Who is covered by health insurance schemes and which services are used in Tanzania? Policy Brief. Shield Health Financing Reform; 2011. 43. Kahabuka C, Moland KM, Kvale G, Hinderaker SG: Unfulfilled expectations to services offered at primary health care facilities: experiences of caretakers of under-five children in rural Tanzania. BMC Health Serv Res 2012, 12:158. 44. Kruk ME, Mbaruku G, McCord CW, Moran M, Rockers PC, Galea S: Bypassing primary care facilities for childbirth: a population-based study in rural Tanzania. Health Policy Plann 2009, 24:279–288. 45. Hidayat B, Thabrany H, Dong H, Sauerborn R: The effects of mandatory health insurance on equity of access to outpatient care in Indonesia. Health Policy Plann 2004, 19(5):322–335. 46. EQUINET: Challenging inequity through redistributive health systems.In Regional Equity Watch 2012: Assessing progress towards equity in health in East and Southern Africa. Harare: EQUINET; 2012. 47. McIntyre D: Health service financing for universal coverage in east and southern Africa.In EQUINET Discussion Paper. No. 95. Harare: EQUINET. 48. Aday L, Andersen R: A framework for the study of access in medical care. Health Serv Res 1974, 9:208–220. 49. Hsiao WC, Shaw PR: Social health insurance in developing countries. Washington DC: World Bank; 2007. 50. Normand C, Weber A: Social health insurance: a guidebook for planning. Submit your next manuscript to BioMed Central Geneva: World Health Organisation; 2005. and take full advantage of: 51. Carrin C: Social health insurance in developing countries: a continuing challenge. Int Soc Security Rev 2002, 55(2):57–69. 52. Stuckler D, Feigl AB, Basu S, McKee M: The political economy of universal • Convenient online submission health coverage. Background Paper for the Global Symposium on Health • Thorough peer review Systems Research. Montreux; 2010. • No space constraints or color figure charges doi:10.1186/1475-9276-13-25 • Immediate publication on acceptance Cite this article as: Chomi et al.: Health care seeking behaviour and • Inclusion in PubMed, CAS, Scopus and Google Scholar utilisation in a multiple health insurance system: does insurance affiliation matter? International Journal for Equity in Health 2014 13:25. • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png International Journal for Equity in Health Springer Journals

Health care seeking behaviour and utilisation in a multiple health insurance system: does insurance affiliation matter?

Loading next page...
 
/lp/springer-journals/health-care-seeking-behaviour-and-utilisation-in-a-multiple-health-nil7p9qA8v

References (81)

Publisher
Springer Journals
Copyright
Copyright © 2014 by Chomi et al.; licensee BioMed Central Ltd.
Subject
Medicine & Public Health; Public Health; Development Economics; Quality of Life Research; Social Policy
eISSN
1475-9276
DOI
10.1186/1475-9276-13-25
pmid
24645876
Publisher site
See Article on Publisher Site

Abstract

Background: Many countries striving to achieve universal health insurance coverage have done so by means of multiple health insurance funds covering different population groups. However, existence of multiple health insurance funds may also cause variation in access to health care, due to the differential revenue raising capacities and benefit packages offered by the various funds resulting in inequity and inefficiency within the health system. This paper examines how the existence of multiple health insurance funds affects health care seeking behaviour and utilisation among members of the Community Health Fund, the National Health Insurance Fund and non-members in two districts in Tanzania. Methods: Using household survey data collected in 2011 with a sample of 3290 individuals, the study uses a multinomial logit model to examine the influence of predisposing, enabling and need characteristics on the probability of seeking care and choice of provider. Results: Generally, health insurance is found to increase the probability of seeking care and reduce delays. However, the probability, timing of seeking care and choice of provider varies across the CHF and NHIF members. Conclusions: Reducing fragmentation is necessary to provide opportunities for redistribution and to promote equity in utilisation of health services. Improvement in the delivery of services is crucial for achievement of improved access and financial protection and for increased enrolment into the CHF, which is essential for broadening redistribution and cross-subsidisation to promote equity. Keywords: Health insurance, Health care utilisation, Health seeking behaviour, Equity Background which to achieve universal coverage given the con- Health insurance has emerged as a key instrument in straints of enforcing universal mandatory coverage [2]. current health financing reforms of middle and low in- However, when the risk pools are fragmented, this also come countries aimed at achieving universal coverage. causes variation in the potential access, health care seek- The health insurance systems of these countries are ing behaviour and utilisation of health services. This is often characterised by multiple health insurance funds likely because apart from reducing the financial barriers covering different population groups. When mechanisms associated with the cost of health services, health insur- to promote cross-subsidies across funds exist within the ance also influences health care seeking behaviour health insurance system, the risk pools are referred to as (whether, when, from where care is sought for an illness) integrated. Without such mechanisms the risk pools are by preventing delays, self-treatment and use of alterna- referred to as fragmented [1]. Arguably, using multiple tive forms of care [3]. In addition, structural features of health insurance funds is the most practical means with the health insurance system such as contribution levels, eligibility and benefit entitlements determine who is cov- * Correspondence: e_chomi@yahoo.com ered as well as the quality, type and quantity of services Equal contributors covered. Hence the way health insurance system is orga- Muhimbili University of Health and Allied Sciences, Dar-es-Salaam, Tanzania Aarhus University, Aarhus, Denmark nised is likely to influence health care seeking behaviour Full list of author information is available at the end of the article © 2014 Chomi et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 2 of 11 http://www.equityhealthj.com/content/13/1/25 and utilisation of health services. Also, the differential Against this background, this paper intends to exam- revenue raising capacities and benefit packages offered ine the effects of fragmented risk pooling on health care by the various insurance funds in a fragmented system seeking behaviour and utilisation of CHF and NHIF are likely to result in varying degrees of access, health members and non-members in two districts in Tanzania. care seeking behaviour and utilisation of health services Specifically we aim to examine the differences in health [1,4]. Furthermore, fragmentation results in inefficiently care seeking behaviour and utilisation between CHF, high administrative costs, which may have an impact on NHIF and non-members. Bivariate and multivariate ana- the ability of the health system to achieve its policy ob- lyses are used to study the relationship between mem- jectives of financial protection [5,6]. bership status and the decision and timing to seek care, Existence of fragmentation also results in a tiered health and choice of provider given a set of predisposing, enab- system, which is inequitable [2]. The experience of many ling and need characteristics. The remainder of this sec- Latin American countries exemplifies this, where social tion provides a discussion of a framework for health care health insurance for formal workers co-existed with na- utilisation. Section two describes the methods and the tional health services delivered directly through ministries data used in this study and section three presents the re- of health to provide the poor and informal workers with sults. Section four discusses the results and the final sec- health service coverage. This led to a tiered system whereby tion provides some conclusions of the study in terms of formal sector workers enjoyed access to a wide range of policy and further analysis. high quality services while the rest had access to a less gen- erous benefit package while incurring higher costs due to Health care seeking behaviour and utilisation co-payments or excluded services [7-10]. Thus, health in- One of the most frequently used frameworks for the surance systems with fragmented risk pools lack the neces- analysis of health care utilisation is Andersen’s behav- sary conditions for cross-subsidisation to promote financial ioural model of health care use [15-21]. This framework protection and equitable utilisation of health services [11]. assumes that utilisation of health care is influenced by Tanzania has two predominant health insurance funds, the predisposition, the ability and the need to use health the Community Health Fund (CHF) and the National services [22,23]. Health Insurance Fund (NHIF). The NHIF is mandatory Predisposing factors relate to the propensity to utilise for public sector employees covering 7.2% of the population health services and include individual characteristics that while the CHF is voluntary and district based for the rural are not directly related to health care utilisation but population with coverage of about 6.6% [12]. Other insur- rather influence the likelihood of utilisation. These char- ance funds include an urban equivalent of CHF for the in- acteristics can be categorised as: demographic, social formal population, ‘Tiba Kwa Kadi’ (TIKA) and the Social structure and health beliefs [24]. Demographic charac- Health Insurance Benefit (SHIB) for members of the Na- teristics include age and sex, which represent biological tional Social Security Fund (NSSF). There are also various factors that affect the likelihood that an individual will private health insurance funds mostly covering those in the need health services. Social structure represents the fac- formal sector through their employers and micro-insurance tors that determine the status of an individual in the so- schemes which cover mostly informal sector workers ciety as well as the physical and social environment. The [13,14]. most common measures of social structure are educa- The health insurance system is fragmented, with no tion level, occupation and ethnicity. Health beliefs are transfers between the risk pools despite the differential the attitudes, values and knowledge that an individual health care needs and revenue bases. NHIF members are may have about health and health services that may in- entitled to a relatively comprehensive package of health ser- fluence utilisation of health services [23]. vices which include specialised services that can be Enabling characteristics describe the means that indi- accessed from government and accredited private primary, viduals have at their disposal with which to utilise health secondary and tertiary care providers. In contrast CHF services. This is based on the argument that even though members are entitled to a package of health care services an individual may be predisposed to utilise health ser- which they can access from primary care providers [13]. vices, certain factors must be in place to enable actual This implies that members of CHF and NHIF have varying use. These include income, health insurance status and degrees of access to health care hence it is likely that this availability of health services. Usually residence (urban/ may influence their health care seeking behaviour and util- rural) and distance are used as proxy measures for avail- isation. While some degree of cross-subsidisation between ability of health services. Need characteristics are the the healthy and the sick occurs within the CHF and NHIF, direct determinants of health care use which include self the lack of cross-subsidies across the funds limits the extent reported and evaluated morbidity [22]. to which resources can be redistributed to promote equit- Health insurance is the primary variable of interest in able utilisation. this study, due to the key role in improving access to Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 3 of 11 http://www.equityhealthj.com/content/13/1/25 health care by reducing financial barriers to utilisation of This approach was used since there are few NHIF mem- health services. It is therefore expected that health insur- bers. Non-member households were randomly selected ance should positively influence the probability of util- from the village household register in each of the se- isation. The fragmentation of the health insurance lected villages. All CHF and NHIF households were system and the differential benefit packages between the omitted from the village register using the list obtained CHF and NHIF implies that members of these two funds from the facility and District Council respectively before will have differential access to health care, which will in- selection of non-member households. In each household herently influence their choice of provider. From this all members were interviewed. perspective, we expect that choice of provider will be in- Data used for analysis was collected as part of a larger fluenced by insurance affiliation. Predisposing and need study, hence estimated sample size calculations were variables are used as control variables. based on all the key study variables and the maximum sample size estimate was used, since this would be suffi- Methodology cient for the analysis of all key variables [27]. Hence, Study setting, design and data collection while we obtained estimates based on the key variables Data for this study was obtained from Kongwa and for this analysis, they were not sufficient for the analysis Mpwapwa districts in Tanzania over a period of eight of other key variables. The sample size calculation was weeks between July and September 2011. The two dis- based on the assumptions that there would be 80% tricts were selected due to their different levels of CHF power to detect a 25% difference between CHF and enrollment, and for convenience in terms of logistics NHIF households in the number of facility visits per year. and costs. Kongwa has a total of 63,612 households We used the proportion from a similar study, which of which 5,800 (9%) are registered with CHF [25]. reported about 50% of insured households having at least Mpwapwa has a total of 78,812 households of which one facility visit per year [28] and assumed the propor- 15,540 (18%) are registered with CHF [25]. The prime tion of NHIF households to be higher. This resulted in a economic activity in both districts is agriculture and live- sample of 729 households (243 per group). Estimating a stock keeping. non-response of 5%, final sample size was adjusted to 766 households. Using the average household size of 5 Sampling method and sample size calculation persons reported by the 2010 TDHS, this represents a For the purposes of this study a household is defined as a sample size of approximately 3830 individuals. person or group of people related or unrelated who live to- A pre-tested structured questionnaire was adminis- gether and share a common pot of food and who share the tered to the household head or spouse. Data was col- same membership card (for CHF) or are dependents of the lected on demographic characteristics, employment, same principal member (NHIF households). This was education level, family size, membership status, house- adapted from the 2010 Tanzania Demographic and Health hold ownership of assets and consumer durables, pres- Survey (TDHS) definition of a household [26]. The study ence of chronic and acute illnesses, general health population comprised of all individuals living in the house- status, health care seeking behaviour and utilisation of holds in the two districts which met this definition. In each health services. Three return visits were made to house- district a multi-stage sampling approach was used to select holds where members were not available for interview first wards, then villages followed by hamlets and eventu- during the first visit, resulting in a response rate of 85%, ally households. Due to difficulties in identification of with a sample size of 3290 individuals from 695 households by membership status from the village house- households. hold register , equal numbers of households were selected from listings of each membership category as follows: CHF Study variables households were randomly selected from the CHF register For the bivariate analysis, the decision and timing of seek- book kept in the health facilities in each ward. This was be- ing care was compared across the CHF, NHIF and non- cause health facilities are registration points for CHF regis- members. The decision to seek care relates to whether or tration. The health facilities were selected based on not care was sought for an illness experience during the whether the facility catchment area falls within the selected four weeks recall period. Timing relates to the time elapsed hamlets. The selection was made from members registered between the onset of symptoms of illness and seeking care from September 2010 to September 2011, to ensure only (same day, less than 1 week, more than 1 week). current CHF members were included. For the multivariate analysis, choice of provider, de- For NHIF households, a list of all Government institu- fined as the place of first contact following an illness tions in the selected wards or villages was obtained from during the four weeks recall period (public hospital, pri- the District Council, from which all available (at the time vate health facility, public health centre/dispensary or of the study) NHIF principal members were selected. traditional healer/self medication) was the dependent Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 4 of 11 http://www.equityhealthj.com/content/13/1/25 variable. The alternative of traditional healer or self The MNL model assumes that the odds of choosing medication refers to those who sought treatment outside between two alternative choices do not depend on which the home from a traditional healer, drugstore or phar- other choices are available (the Independence of Irrele- macy. This differs from individuals who did not seek vant Alternatives (IIA) assumption) [33]. We employed care but instead used home remedies or those who de- Hausman-McFadden (HM) and Small-Hsiao (SH) tests layed seeking care and opted to start with home remed- to validate this assumption. Both tests returned non- ies first. Traditional healer and self medication was later significant results (HM-ρ=0.112; SH-ρ=0.112), indicating merged since both choices represent alternative sources that the model is appropriate. The use of alternative of care. In addition the choice of traditional healer models that relax the IIA assumption such as the multi- accounted for only 1% of those seeking care. nomial probit, nested logit and mixed logit models is Drawing on Anderson’s 1995 model [23], independent limited by their computational difficulties and for the variables used in this study include predisposing, enabling multinomial probit, the need for a particular data struc- and need characteristics of individuals. Predisposing charac- ture [32]. teristics included age (0–5, 6–14, 15–49, 50–59, 60+ years), It is also possible that the same factors that influence sex (male, female) and education level of household head health care utilisation could also influence the purchase (no education, primary education, secondary education, of health insurance, implying that the health insurance above secondary education). Perceived adequacy of staff variable is endogenous [33-35]. The mandatory nature of and perceived availability of drugs (yes, no) were included NHIF and household basis of membership for the CHF as proxies for attitudes towards health services as expressed reduce the effect of selection bias in our study. However, from general questions on health status and utilisation. possible endogeneity of the health insurance variable Enabling characteristics included were household char- was tested using the Durbin-Wu-Hausman test, which acteristics that were assigned to an individual according uses an instrumental variable to test whether the predic- to the household to which he/she belonged. These were tors are correlated with the error term. A non- membership status (CHF, NHIF, non-members), resi- significant test result indicates that none of the predictor dence (urban, rural), distance to facility (less than or variables are endogenous [36]. We used relationship to more than 5 km) and wealth status (lowest to highest the head of household as the instrumental variable and wealth quintile). Owing to the complexities of determin- the test was not significant (ρ=0.315) implying that the ing actual income, Principal Components Analysis health insurance variable was exogenous. Choice of the (PCA) was used to develop an asset index that grouped instrumental variable was based on its influence on households into quintiles based on ownership of assets health insurance membership but not on health care and durable goods [29,30]. The asset index of the house- utilisation, which is a criterion for instrumental variable hold was used to represent socio-economic status. For selection. Being related to the household head makes the need characteristics we used self reported illness, de- one eligible for insurance membership, but does not in- fined as the experience of illness or injury lasting for a fluence whether or where care will be sought for an month or less (acute) or experience of an illness lasting illness. for three months or more (chronic). Since individuals in the sample were obtained from households, the observations of each individual are not Analysis independent of each other, resulting in an under- Bivariate and multivariate analysis was used to study the re- estimation of standard errors and making significance lationship of predisposing, enabling and need characteris- tests used in the analysis invalid. The effect of clustering tics and health care seeking behaviour and utilisation. Chi has been accounted for in the analysis, using clustered square tests were used to study the relationship between robust standard errors which increase the variability be- membership status and health care seeking behaviour along tween individuals within cluster [37-39]. two dimensions: the decision to seek care for an illness and timing. Multinomial logistic (MNL) regression was used to Ethical considerations estimate the choice of provider given a set of predisposing, Ethical approval was sought from the Research and Ethics enabling and need characteristics. This model was selected Committee of Muhimbili University of Health and Allied based on the nature of the dependent variable and the abil- Sciences. Following ethical approval, permission to con- ity of the model to estimate all choices in a single equation duct data collection was obtained from the Regional [31,32]. Since the aim was to find out whether insured indi- Administrative Secretary (RAS) of Dodoma and District viduals will choose a provider where they can use their in- Administrative Secretary (DAS) of Kongwa and Mpwapwa. surance card, the health centre/dispensary category was Respondents were informed of the research objectives and used as a reference (both CHF and NHIF members can use were asked to participate in the study. Those who agreed their cards at this level). were asked to sign a consent form. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 5 of 11 http://www.equityhealthj.com/content/13/1/25 Results education, while the majority of CHF (72%) and non- Descriptive characteristics of the sample members (64%) attained primary education (ρ<0.05). Table 1 presents the descriptive characteristics of the The majority of NHIF households are relatively wealthy study sample. About 28% of the respondents were from (39% highest, 40% fourth wealth quintile) compared to NHIF households, 38% from CHF households and 34% CHF (29% lowest, 31% second wealth quintile) and non- from non-member households. More than 50% of NHIF member households (33% lowest, 27% second wealth members were in the 15–49 year age group, with less quintile, ρ<0.05). than 10% aged 0–5 years and less than 3% aged 60 years and above. In contrast 43% each of CHF and non- Health care seeking behaviour and utilisation members were aged 15–49 years, 15% aged 0–5 years Table 2 illustrates health care seeking behaviour by and 4% aged 60 and above (ρ<0.05). A higher proportion membership status. During the recall period of four of CHF households (44%) had more than five members, weeks prior to the survey, 30 percent reported having compared to NHIF (28%) and non-member households had at least one spell of illness. Of those individuals, 75 (31%, ρ<0.05). percent reported having sought care for their illness. NHIF head of households were more educated, with 63% having attained secondary or above secondary Delays in seeking care The majority of CHF (57%) and NHIF (66%) members were more likely to seek care on the same day they fell Table 1 Individual characteristics by membership status, ill, while more than 40% of non-members were more Kongwa and Mpwapwa 2011 (%) likely to experience delays in seeking care (ρ<0.05 ). Variable NHIF CHF Non-members The reasons for delaying to seek care and/or not seek- N = 931 N = 1242 1117 ing care are presented in Table 3. Among those who de- Age*** layed seeking care, the most common reason reported 0-5 9.1 14.5 14.8 was “wait and see if illness progresses further” for NHIF 6-14 25.2 33.4 31.3 (40%) and CHF (44%) members. Among non-members 15-49 55.2 42.9 42.9 lack of money to pay for treatment was the main reason, reported by 34%. Reasons for not seeking care differed 50-59 8.7 4.9 6.5 significantly by membership status. The main reasons 60+ 1.8 4.3 4.4 were availability of own medicine/home remedies, re- Sex ported by 43% of NHIF and 38% of CHF members and Male 47.1 48.6 46.9 lack of money to pay for treatment, reported by 51% of Female 52.9 51.5 53.1 non-members (ρ<0.05). Education (head)*** Choice of Provider No education 1.4 20.7 30.7 The most common choice of provider was the public Up to Primary 8.1 71.9 63.9 health centre/dispensary; 43% of all individuals seeking Up to–Secondary 27.9 6.5 5.0 care chose this source, followed by traditional healer/self Above secondary 62.6 0.7 0.4 medication (28%), public hospital (18%) and private Household size*** Mean = 4.7 4.0 5.2 4.7 Table 2 Health seeking behaviour by membership status, 1-5 members 72.3 56.2 69.3 conditioned on reporting illness (%) >5 members 27.7 43.8 30.7 Variable NHIF CHF Non-members Wealth Status*** N = 931 N = 1242 N = 1117 Lowest 2.5 29.4 33.2 Reported illness (yes) 31.3 30.5 28.6 Second 3.7 30.8 27.2 Sought care for illness?*** N = 291 N = 379 N = 320 Third 14.7 22.7 22.7 Yes 81.1 77.3 65.6 Fourth 40.2 12.7 10.3 No 18.9 22.7 34.4 Highest 38.9 4.4 6.5 Timing of care** N = 236 N = 293 N = 201 Distance to facility Same day 66.1 57.3 52.7 1-5 km 91.9 95.4 95.4 <1 week 31.4 37.8 40.8 >5 km 4.6 8.1 4.7 >1 week 2.5 4.8 6.5 ***ρ<0.01; based on Chi square test. **ρ<0.05; ***ρ<0.01; based on Chi square test. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 6 of 11 http://www.equityhealthj.com/content/13/1/25 Table 3 Reasons for delays and/or not seeking care by dispensaries, compared to 28% of NHIF members and membership status, conditioned on reporting illness (%) 35% non-members (ρ<0.05). Private facilities were pre- Variable NHIF CHF Non-members ferred by 22% of NHIF members, 5% of CHF members and 10% of non-members. Reasons for delays*** N = 77 N = 120 N = 90 No money 9.1 9.2 40.0 Reasons for choice of provider Distance 6.5 1.7 1.1 Reasons for choice of provider are categorised into those Self treatment (home) 18.2 16.7 12.2 for choice of formal care and those for choice of alterna- “Wait and see” 49.4 49.2 33.3 tive forms of care (Table 4). Good quality (availability of Facility closed 10.4 11.7 10.0 drugs, laboratory tests, staff and most likely to find a doctor) was the main reason for seeking care from a dis- Other 11.7 11.7 3.3 trict hospital, reported by 41% of non-members and Reasons for not seeking care*** N = 51 N = 68 N = 97 NHIF members who used this provider. In contrast, 52% Poor quality of services 11.8 1.5 0.0 of CHF members who sought care from a public hospital No money 5.9 10.3 50.5 said it was the only facility nearby (ρ<0.05). Quality rea- No insurance card 1.9 0.0 5.2 sons were also the most important for those who sought Distance 0.0 1.5 0.0 care from a private facility, reported by 73% of NHIF members, 50% of CHF members and 65% of non- Illness not serious 23.5 30.9 11.3 members (ρ<0.05). No one to leave at home/farm 3.9 10.3 1.0 The main reason for seeking care from public pri- Had my own medicine/home 43.1 38.2 26.8 mary care facilities is ‘only facility available nearby’, remedies reported by 61% of NHIF members, 64% of CHF Other 7.4 9.8 5.2 members and 75% of non-members who sought care ***ρ<0.01; based on Chi square test. from this source. Of particular interest is the import- facilities (12%). About 22% of NHIF members who ance of insurance for CHF members, reflected by sought care for an illness went to a public hospital, com- about 30% whose choice of provider was based on pared to 16% of CHF members and 16% of non- the ability to use their insurance card, compared to members. A higher proportion of non-members opted 20% and 18% of NHIF members. for traditional healers/self medication (40%), compared Among those who sought alternative forms of care, to that of NHIF members (29%) and CHF members the main reason reported was that they were more likely (20%, ρ<0.05). Figure 1 illustrates the different choice to receive treatment from this source rather than from patterns by membership status. It shows that 60% of formal sources of care (NHIF 43%, CHF 47%, non- CHF members sought care from public health centre/ members 27%, ρ<0.05). CHF NHIF non member choice of provider public hospital private hospital/HC/D/C public HC/D traditional/ self medication Figure 1 Choice of provider by membership status, conditioned on reporting illness (%). 0 20 40 60 Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 7 of 11 http://www.equityhealthj.com/content/13/1/25 Table 4 Reasons for choice of care by membership status healer/self medication over a public health centre/dispens- and provider, conditioned on seeking care (%) ary. Urban residents are more likely than rural residents to Variable NHIF CHF Non-members choose a public hospital, private facility and self medication N = 171 N = 237 N = 128 over a public health centre/dispensary, regardless of health District hospital (N = 130)*** insurance status. Individuals are more likely to travel long Good quality 40.8 15.2 41.3 distances to seek care from a public hospital rather than a public health centre/dispensary. Only facility available nearby 34.7 52.2 41.2 Insurance card accepted 20.1 30.4 NA The effect of need characteristics Exemption NA NA 8.8 Individuals reporting chronicillness aremorelikelyto Private Health centre/Dispensary (N = 85)*** choose a public hospital or a private facility over a public Good quality 72.6 50.0 65.0 health centre/dispensary. Results indicate that acute illness Only facility available nearby 17.7 42.9 30.0 is not a significant determinant of choice of provider. Insurance card accepted 5.9 0.0 NA Discussion Exemption NA NA 0.0 This paper examined the effects of fragmented risk pool- Public Health center/Dispensary (N = 315)*** ing on health care seeking behaviour and utilisation of Good quality 21.2 5.1 17.8 CHF and NHIF members and non-members. Results Only facility available nearby 60.6 64.2 75.3 suggest that the insured are more likely to seek care and Insurance card accepted 18.2 30.6 NA less likely to experience delays compared to non- members. Lack of money to pay for treatment is a sig- Exemption NA NA 2.7 nificant barrier to seeking care for non-members but not Choice of alternative care (traditional healer/self for CHF and NHIF members, since this was the main rea- medication, N = 209)** son for delays in seeking care or not seeking care at all N = 58 N = 51 N = 70 reported by non-members. This suggests that generally No money 5.2 11.8 21.4 health insurance does improve access to health care by re- Distance 17.2 13.7 8.6 ducing the financial barriers associated with utilisation of More likely to get treatment 43.1 47.1 27.1 health services and is consistent with results reported by Poor quality of services at 8.6 7.8 12.9 Jutting [37], Bronwyn et. al [40], Mensah et al. [41] and formal facility SHIELD [42]. However, members of the NHIF are more No insurance card 0.0 3.9 7.1 likely than CHF members to seek care for an illness and Illness not serious 8.6 3.9 2.7 are also less likely to delay seeking care. This variation cor- roborates findings of Ekman [4] in Jordan who reported a Recognise symptoms, familiar 6.9 3.9 14.3 with drugs higher probability of seeking care among members of the Ministry of health insurance programme compared to Other 8.6 7.8 5.7 other programmes. The variation between CHF and NHIF **ρ<0.05; ***ρ<0.01; based on Chi square test. Percentages may not add up to 100% since only the 3 main reasons members can be explained by the fact that compared to are presented. CHF members, NHIF members are more likely to live near a health facility and have a wider choice of providers Multivariate results compared to CHF members. Table 5 presents results from the multinomial logit Differences were also found in relation to the choice model specification. of provider. Compared to CHF members, NHIF mem- bers are more likely to choose a private facility, trad- The effect of enabling characteristics itional healer or self medication over a public health Persons from wealthier households are more likely to centre or dispensary. Although not significant, results choose a public hospital, private facility and traditional also show that NHIF members are more likely than CHF healer/self medication over a public health centre/dispens- members to choose a public hospital rather than a pub- ary compared to those from relatively poorer households. lic health centre or dispensary. Good quality (availability The positive sign on the coefficient indicates that compared of drugs, laboratory tests, staff and most likely to find a to CHF members, NHIF members are more likely to doctor) was the main reason for seeking care from a dis- choose a private facility or traditional healer/self medication trict hospital and private facility. District hospitals and over a public health centre/dispensary. Likewise, non- private facilities are usually relatively better equipped members are more likely than CHF members to choose a and are more likely to have drugs than dispensaries or public hospital, private health facility and a traditional health centres [25]. Moreover, a non-functional referral Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 8 of 11 http://www.equityhealthj.com/content/13/1/25 Table 5 Multinomial logit estimation results, (public health centre/dispensary as reference) Variable Public hospital Private hospital Traditional/self Coeff. Robust Std Error Coeff. Robust Std Error Coeff. Robust Std Error Age (0–5 ) 6-14 −0.241 0.369 0.113 0.437 0.340 0.302 15-49 0.116 0.340 0.693* 0.413 0.836** 0.304 50-69 0.707 0.548 0.131 0.737 1.010** 0.465 60+ −1.367 1.043 −0.363 1.103 0.648 0.579 Sex (male ) Female 0.023 0.260 0.250 0.299 −0.142 0.200 Education level (no education ) Up to Primary 0.437 0.540 0.378 0.638 0.480 0.405 Up to Secondary −0.224 0.702 −0.964 0.908 −1.766** 0.646 Above Secondary −1.346 0.567 −1.311 0.940 −2.191*** 0.716 Wealth status (lowest ) Second −0.533 0.594 −0.071 0.651 0.239 0.377 Third 0.360 0.490 0.214 0.693 0.811** 0.357 Fourth 1.008 0.626 1.295* 0.728 1.525** 0.553 Highest 1.675** 0.657 1.940** 0.746 1.313** 0.590 Membership status (CHF ) NHIF 0.037 0.679 1.664** 0.770 2.123*** 0.574 Non-member 1.078** 0.411 1.594*** 0.506 1.520*** 0.307 Residence (rural ) Urban 3.721*** 0.453 1.868*** 0.446 1.046*** 0.359 Distance 0.091 0.072 0.075 0.070 0.041 0.062 Self reported morbidity (no illness ) Acute 0.243 0.675 0.773 0.912 0.076 0.548 Chronic 1.695** 0.821 2.201** 1.022 0.980 0.677 Perceived quality of care (poor ) Availability of drugs −0.791** 0.374 −1.085*** 0.356 −0.358* 0.279 Adequacy of staff 0.196 0.371 0.388 0.326 0.386 0.300 Constant −3.931*** 0.894 −4.762*** 1.380 −2.620*** 0.733 *ρ<0.1; **ρ<0.05; ***ρ<0.01; base category. system and no gate-keeping makes it easier for patients ‘insurance card accepted’ (31%), while quality reasons to bypass primary care facilities to seek care from hospi- were reported by only 5% of CHF members. This implies tals [43,44]. Given that the NHIF coverage extends to all that CHF members will choose a provider that is nearby levels of care and to private accredited facilities, it is not and where they can use their insurance card to pay for surprising that members chose providers on the basis of services regardless of quality of care offered. The CHF perceived better quality of care. A study in Indonesia benefit package is limited to primary level care offered at examining the effects of mandatory insurance on access health centres and dispensaries, and CHF registration to carealsofound apositiveeffect of theAskes (for occurs at the health facility, usually one that is nearest public employees) on the use of public providers, while to where the member lives. Hence it is possible that the Jamostek (for private employees) had a positive ef- members link their insurance entitlements to the facility fect on useof bothpublicand privateproviders.This where they registered making this facility the most lo- was attributed to the differential benefit packages by the gical choice. Therefore, unlike NHIF members, for CHF two funds [45]. members, choice of provider involves a trade-off be- The main reasons for choice of provider among CHF tween claiming their entitlements from providers who members were ‘only facility available nearby’ (64%) and may not necessarily provide the desired quality of care Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 9 of 11 http://www.equityhealthj.com/content/13/1/25 and forgoing their entitlements by paying out-of pocket In the Andersen behavioural model of health care util- for perceived superior quality. This may make health in- isation, enabling characteristics such as income and surance an unattractive product, reducing the likelihood health insurance are described as those that are neces- of re-enrolment of CHF members and negatively influ- sary but not sufficient for utilisation [48]. Adequate ence non-members’ decision to join the fund. health infrastructure capable of delivering quality health The variation in health seeking behaviour and choice services has been mentioned as one of the pre-requisites of provider between the CHF and NHIF members is a of successful implementation of health insurance [49-51] reflection of the effects of a fragmented health insurance and achievement of a truly universal health system [52]. system. The lack of a standardised benefit package for The same arguments have been raised by Robyn et. al the members of the two funds implies that while both [3], who found that the effect of health insurance on CHF and NHIF members have better potential access, health seeking behaviour in Burkina Faso was limited, the quality and quantity entitled to respective members owing to poor quality of services. This means that differs. In other words the whether, when, where health achieving basic coverage of the population is meaning- care is sought and the quality and quantity of health ser- less when this coverage does not guarantee access to ser- vices received depends on health insurance affiliation. vices of an adequate quality. Our findings show that the Given the lower revenue raising capacity of CHF com- extent to which health insurance promotes health care pared to the NHIF, it also means that access to health utilisation is dependent on the quality of services offered care is based on ability to pay rather than on need for at formal care providers. When patient expectations are the services. This goes against the principle of equity of not met, it is likely that they will seek alternatives even access and calls for mechanisms that will promote though it means they have to forgo their insurance bene- broader risk sharing and redistribution across the two fits. This has been elucidated by our results, which show schemes and a standardised benefit package. In this way that the majority of CHF and NHIF members who for a given need, insured individuals will enjoy the same sought care from private facilities and from alternative degree of access to health services regardless of insur- sources (self-medication/traditional healer) did so due to ance affiliation or ability to pay. perceived inferior quality in public facilities. Given the One of the goals of expanding health insurance in the voluntary nature of CHF, and its potential for covering Tanzanian health system was to eventually achieve uni- the majority of the population, quality improvements in versal coverage and universal access to health care. For a health services are important to encourage enrolment. system to be truly universal it has to be equitable, grant- Scaling up enrolment into CHF is important for the re- ing access to the same range of services for all based on distributive potential of the scheme and ultimately of an need while requiring payment for these services based integrated health insurance system. on their income [46,47]. This can only be achieved in a health insurance system that is redistributive, such that Conclusion there is risk sharing and cross-subsidisation across insur- This paper examined the effects of fragmented risk pool- ance schemes. Expanding health insurance to cover all ing on health care seeking behaviour and utilisation of population groups without redistributive mechanisms may CHF and NHIF members and non-members. Specific achieve universal coverage but also create a system that areas that were examined were the decision to seek care does not support equity. Our results have shown the in- when ill and the timing and choice of health service pro- equalities in access between the CHF and NHIF members. viders. The findings of this study provide lessons for pol- The expansion of health insurance is a reflection of icy makers in low- and middle-income countries where commitment by the Tanzanian government to achieve multiple health insurance funds have been established to universal coverage. What remains is the development of achieve universal coverage. In particular, addressing the policy framework and design issues that will promote re- challenges of limited risk sharing and cross-subsidisation distribution and cross-subsidisation across the schemes in across multiple health insurance funds remains crucial order to create a health insurance system that is universal for equitable access. These results confirm the import- and equitable. Creating a standardised benefit package is a ance of reducing fragmentation in risk pooling arrange- crucial step in promoting equitable access to health ser- ments, creating the opportunity for risk and income vices. The requirement for a standardised benefit package cross-subsidisation that will also promote the develop- lies in the link between the package, risk structure and ex- ment of a standard benefit package. pected health expenditures of a scheme. Without a stan- Perceived poor quality of health limits the degree to dardised benefit package, redistribution across schemes which the objectives of improved access and financial will reward those with more comprehensive packages at protection can be achieved. Furthermore, poor quality of the expense of schemes with fewer benefits, perpetuating health services serves as a deterrent for enrolment into rather than reducing inequity. voluntary funds, which often represent crucial elements Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 10 of 11 http://www.equityhealthj.com/content/13/1/25 for broadening redistribution and cross-subsidisation to insurance on reducing informal self-care in Burkina Faso. Health Policy Plann 2011, 27(2):156–165. promote equity. 4. Ekman B: The impact of health insurance on outpatient utilisation and expenditure: evidence from one middle-income country using national household survey data. Health Res Policy Syst 2007, 5:6–21. Endnotes 5. Kirigia JM, Preker G, Mwikisa C, Diarra-Nama AJ: An overview of health Principal member is the contributing member of the financing patterns and the way forward in the WHO African Region. NHIF, usually the head of household or spouse. East Afr Med J 2006, 83(8):S1–S28. 6. McIntyre D: Learning from experience: health care financing in low- and In Tanzania, districts are organised into divisions, middle-income countries. Geneva: Global Forum for Health Research; 2006. which in turn are divided into wards. Within each ward, 7. Wagstaff A: Social health insurance re-examined. World Bank Policy Research there are a number of villages, which are also divided Working Paper 4111. Washington DC: World Bank; 2007. into hamlets. Depending on the ward and health infra- 8. Pinto D, Hsiao WC: Colombia: Social health insurance with managed competition to improve health care delivery.In Health systems in low- and structure, one health facility may have a catchment area middle-income countries. Edited by Hsiao WC, Shaw PR. Washington DC: of one or more villages. World Bank; 2007:105–132. Each village has a list of all households registered at 9. Baeza CC, Packard TG: Beyond survival: protecting households from health shocks in Latin America. California: Stanford University Press: Washington DC: the office of the Village Executive Officer. This list is World Bank; 2006. broken down by hamlet but does not show membership 10. Jack W: Health insurance reform in four Latin American countries: theory status. and practice. Policy Research Working Paper, Volume 2492. Washington DC: World Bank; 2000. 11. McIntyre D: What healthcare financing changes are needed to reach Abbreviations universal coverage in South Africa? S Afr Med J 2012, 102(6):489–490. CHF: Community Health Fund; NHIF: National Health Insurance Fund; TIKA: Tiba kwa Kadi (urban equivalent of CHF); SHIB: Social Health Insurance 12. National Health Insurance Fund: Actuarial valuation report of the National Benefit; NSSF: National Social Security Fund; TDHS: Tanzania Demographic Health Insurance Fund. Dar-es-Salaam: M/S iPrint Ltd; 2010. and Health Survey; PCA: Principal Components Analysis; MNL: Multinomial 13. Ministry of Health and Social Welfare: Health financing in Tanzania: where logit; IIA: Independence of Irrelevant Assumptions; RAS: Regional we came from and where we are now. Presented at the National Forum Administrative Secretary; DAS: District Administrative Secretary. for Health Financing. Dar-es-Salaam; 2010. 14. Haazen D: Making health financing work for poor people. Washington DC: World Bank; 2012. Competing interests 15. Young JT, Menken J, Williams J, Khan N, Kuhn R: Who receives healthcare? The Authors declare no competing interests. Age and sex differentials in adult use of healthcare services in rural Bangladesh. World health Popul 2006, 8(2):83–100. Authors’ contributions 16. Brown C, Barner J, Bohman T, Richards K: A multivariate test of an ENC was involved in study design, training of research assistants, review of expanded Andersen health care utilisation model for complementary data collection tools, data collection, data entry, data analysis and drafting and alternative medicine (CAM) use in African Americans. J Altern the manuscript, revising and writing of the final manuscript. PGMM, UE, ADK Complem Med 2009, 15(8):911–919. and KH were involved in study design, review of data collection tools, data 17. Lopez-Cevallos DF, Chi C: Assessing the context of health care utilisation in analysis and revising the manuscript. All the authors have read and approved Ecuador: a spatial and multilevel analysis. BMC Health Serv Res 2010, 10:64. the final version of the manuscript. 18. Jahangir E, Irazola V, Rubinstein A: Need, enabling predisposing and behavioural determinants of access to preventive care in Argentina: Acknowledgements analysis of the national survey of risk factors. PLoS ONE 2012, 7(9):e45053. This research was part of a research program on health insurance in Ghana 19. Glei DA, Goldman N, Rodriguez G: Utilisation of care during pregnancy in and Tanzania, supported by the Danish Consultative Committee on rural Guatemala: does obstetrical need matter? Soc Sci Med 2003, Development Research (Grant DFC 19-08-AAU) within the framework of the 57:2447–2463. Danish development cooperation. The funder was not involved in the 20. Nanjin N, Bennett CM, Luby S: Inequalities in care-seeking for febrile design, data collection, analysis and interpretation of data, or decision to illness of under-five children in urban Dhaka, Bangladesh. J Health Popul publish the results. We also appreciate the Muhimbilii University of Health Nutr 2011, 29(5):523–531. and Allied Sciences, Dar es Salaam for providing ethical approval to conduct 21. Kamgnia B: Use of health services in Cameroon. Int J Appl Econometrics the study. In addition we are thankful for the time spared by the participants Quantitative Studies 2006, 3(2):55–64. who contributed to the successful conduct of this study; the District Councils 22. Andersen R, Newman J: Societal and individual determinants of medical of Kongwa and Mpwapwa and the Ward and Village authorities where the care utilization in the United States. Milbank Q 1973, 51(1):95–124. data was collected who granted permission to conduct the study. 23. Andersen R: Revisiting the behavioural model and access to medical care: does it matter? J Health Soc Behav 1995, 36:1–10. Author details Muhimbili University of Health and Allied Sciences, Dar-es-Salaam, Tanzania. 24. Andersen R: A behavioural model of families’ use of health services.In 2 3 Aarhus University, Aarhus, Denmark. London School of Hygiene and Research Series No. 25. Chicago: Center for Health Administration Studies, Tropical Medicine, London, UK. University of Chicago; 1968. 25. United Republic of Tanzania: Tanzania Health Statistical Abstract. Dar-es- Received: 25 September 2013 Accepted: 12 March 2014 Salaam: Ministry of Health and Social Welfare; 2011. Published: 19 March 2014 26. United Republic of Tanzania: Tanzania demographic and health survey 2010. Dar-es-Salaam: National Bureau of Statistics; 2011. 27. Varkevvisser CM, Pathmanathan I, Brownlee A: Designing and Conducting References Health Systems Research Projects. Volume 1: Proposal Development and 1. Smith P, Witter S: Risk pooling in health care financing: the implications for Fieldwork. Amsterdam: KIT Publishers, Ottawa: International Development health system performance. HNP Discussion Paper. Washington DC: World Research Centre. World Health Organisation-Regional Office for Africa: Bank; 2004. Brazzaville; 2003. 2. McIntyre D, Kutzin J: Revenue collection and pooling arrangements in financing.In Health systems in low- and middle-income countries. Edited by 28. Gnawali DP, Pokhrel S, Sie A, Sanon M, De Allegri M, Souares A, Dong H, Smith RD, Hanson K. New York: Oxford University Press; 2012:77–101. Sauerborn R: The effect of community-based health insurance on the 3. Robyn PJ, Hill A, Liu Y, Souares A, Savadogo G, Sié A, Sauerborn R: utilisation of modern health care services: evidence from Burkina Faso. Econometric analysis to evaluate the effect of community-based health Health Policy 2009, 90(2–3):214–222. Chomi et al. International Journal for Equity in Health 2014, 13:25 Page 11 of 11 http://www.equityhealthj.com/content/13/1/25 29. Vyas S, Kumaranayake L: Constructing socio-economic status indices: how to use principal components analysis. Oxford University Press; 2006. 30. Filmer D, Pritchett L: Estimating wealth effects without income or expenditure data-or tears. An application to educational enrolment states in India. World Bank Policy Research Working Paper. Washington DC: World Bank; 2001. 31. Agresti A: An introduction to categorical data analysis. 2nd edition. New Jersey: John Wiley and Sons, Inc.; 2007. 32. Cheng S, Long JS: Testing for IIA in the multinomial logit model. Socio Meth Res 2007, 35(4):583–600. 33. Long JS, Freese J: Regression models for categorical dependent variable using Stata. Texas: Stata Press; 2006. 34. Waters HR: Measuring the impact of health insurance with a correction for selection bias-a case study of Ecuador. Health Econ 1999, 8(5):473–483. 35. Davidson R, MacKinnon JG: Econometric Theory and Method. New York: Oxford University Press, 2004. In Ekman B: The Impact of Health Insurance on Outpatient Utilisation and Expenditure: Evidence from one Middle-income Country using National Household Survey Data. Health Res Policy and Syst 2007, 5:6–21. 36. UCLA: Statistical Consulting Group: Analysing correlated (clustered) data. [http://www.ats.ucla.edu/stat/stata/library/cpsu.htm] 37. Juul S: An introduction to Stata for health researchers.2nd Edition. Texas: Stata Press; 2008. 38. Jowett M, Deolalikar A, Martinsson P: Health insurance and treatment seeking behaviour: evidence from a low-income country. Health Econ 2004, 13:845–857. 39. Jutting JP: Do community-based health insurance schemes improve poor people’s access to health care? Evidence from rural Senegal. World Dev 2004, 32:273–288. 40. Bronwyn H, Goudge J, Ataguba JE, McIntyre D, Nxumalo N, Jikwana S, Chrisich M: Inequities in access to health care in South Africa. J Public Health Policy 2011, 32:S102–S123. 41. Mensah J, Oppong JR, Bobi-Barimah K, Frempong G, Sabi W: An evaluation of the Ghana National Health Insurance Scheme in the context of the health MDGs. Global Development Network Working Paper Series 2010, No. 40. GDN. 42. SHIELD: Who is covered by health insurance schemes and which services are used in Tanzania? Policy Brief. Shield Health Financing Reform; 2011. 43. Kahabuka C, Moland KM, Kvale G, Hinderaker SG: Unfulfilled expectations to services offered at primary health care facilities: experiences of caretakers of under-five children in rural Tanzania. BMC Health Serv Res 2012, 12:158. 44. Kruk ME, Mbaruku G, McCord CW, Moran M, Rockers PC, Galea S: Bypassing primary care facilities for childbirth: a population-based study in rural Tanzania. Health Policy Plann 2009, 24:279–288. 45. Hidayat B, Thabrany H, Dong H, Sauerborn R: The effects of mandatory health insurance on equity of access to outpatient care in Indonesia. Health Policy Plann 2004, 19(5):322–335. 46. EQUINET: Challenging inequity through redistributive health systems.In Regional Equity Watch 2012: Assessing progress towards equity in health in East and Southern Africa. Harare: EQUINET; 2012. 47. McIntyre D: Health service financing for universal coverage in east and southern Africa.In EQUINET Discussion Paper. No. 95. Harare: EQUINET. 48. Aday L, Andersen R: A framework for the study of access in medical care. Health Serv Res 1974, 9:208–220. 49. Hsiao WC, Shaw PR: Social health insurance in developing countries. Washington DC: World Bank; 2007. 50. Normand C, Weber A: Social health insurance: a guidebook for planning. Submit your next manuscript to BioMed Central Geneva: World Health Organisation; 2005. and take full advantage of: 51. Carrin C: Social health insurance in developing countries: a continuing challenge. Int Soc Security Rev 2002, 55(2):57–69. 52. Stuckler D, Feigl AB, Basu S, McKee M: The political economy of universal • Convenient online submission health coverage. Background Paper for the Global Symposium on Health • Thorough peer review Systems Research. Montreux; 2010. • No space constraints or color figure charges doi:10.1186/1475-9276-13-25 • Immediate publication on acceptance Cite this article as: Chomi et al.: Health care seeking behaviour and • Inclusion in PubMed, CAS, Scopus and Google Scholar utilisation in a multiple health insurance system: does insurance affiliation matter? International Journal for Equity in Health 2014 13:25. • Research which is freely available for redistribution Submit your manuscript at www.biomedcentral.com/submit

Journal

International Journal for Equity in HealthSpringer Journals

Published: Mar 19, 2014

There are no references for this article.