Economic burden of non-communicable diseases on households in Nigeria: evidence from the Nigeria living standard survey 2018-19

Data

We conducted a descriptive study using nationally representative secondary cross-sectional data from the most recent round of the Nigerian Living Standard Survey (NLSS) conducted by Nigeria’s National Bureau of Statistics (NBS) between September 2018 and September 2019 [29]. The survey used a stratified multi-stage random sampling to sample 22,200 households with a mean household size of 5.06 and about 116,320 individuals. It covered 600 households per state, including the Federal Capital Territory (FCT), with a 95% response rate. The survey collected individual and household level data, including demographic variables, health problems and access, education, labour, consumption, housing, and assets. We excluded Borno State from our analyses because insecurity prevented the survey of many households in the state.

Diseases and health conditions classification

The health conditions provided in the NLSS 2018/19 were self-reported. Respondents were asked: “Were you sick or injured in the last 30 days?” Those who answered “yes” were asked to list two illnesses/injuries in order of severity. Unlisted conditions were entered under “other (specify)”.

We matched the two lists of health conditions and the “other (specify)” category to ICD-10 classifications to distinguish between NCDs (excluding injuries) and infectious diseases. Using the list provided in the WHO (2013) report [30] and the Gouda et al. (2019) study [4], we arrived at 14 NCD categories: CVDs, cancers (neoplasms), respiratory diseases, diabetes, mental disorders, neurological diseases, haematological diseases (mainly sickle cell diseases), sense organ diseases, renal diseases (kidney), gastroenterological (digestive) diseases, musculoskeletal diseases, dermatological diseases (skin and subcutaneous), dental (oral) diseases and Other NCDs (urinary disorders, gynaecological diseases, and endocrine disorders) (Supplementary Table 1). We grouped injuries under “other illnesses and injuries” because they are not NCDs.

Depending on how they were named in the survey, some NCDs may belong to “other illnesses and injuries” and vice versa. For example, the survey lumped all body pains together, making chronic musculoskeletal pains indistinguishable. Also, for cough, the primary disease conditions, such as COPD, were not specified. Thus, we omitted body pain and cough to reduce bias. Diseases under the names respiratory, tooth, ear, and eye problems in the survey may be either NCDs or infectious diseases. So, they were excluded. The age group covered in our estimations was from five years and above. We excluded the under-five age group to avoid misclassifying infectious diseases commonly found in under-five children in these ambiguous categories [25]. This would not likely underestimate the number of NCDs in our study since the peak age for NCDs in Nigerian children was 6–11 [31], and many children with NCDs don’t live to age 5 in LMICs [32].

Health services utilisation and cost

Respondents that utilised healthcare were asked about the duration of the illness/injury and the amount paid for consultation (excluding drugs). They were further asked about any drugs purchased. The recall period for both was 30 days. In addition, they were asked if they had been admitted to a hospital or health facility in the previous 12 months. Those that answered in the affirmative were asked how long and how much they paid (excluding drugs). Following Mahal et al. (2010) [33], the costs of consultation and drugs were annualised by multiplying them by 12.17, presuming that the visits were uniformly distributed throughout the year. Individual-level data were collapsed and merged with household-level data for household-level estimates.

Consumption and income

We aggregated the consumption data in the survey at the household level. They covered five types of household consumption: food (bought, self-produced, and gifted), non-food (various regularly purchased commodities and services), education, health, and rent. Before statistical analyses, household income and consumption data were examined for missing values, and none was found. Unrealistic negative or no food expenditures data were dropped.

Analytical strategy and statistical analysis

Estimation of household catastrophic health expenditure

Wagstaff and Doorslaer’s budget share [14] and WHO’s capacity-to-pay (CTP) [15] are methods commonly used to estimate CHE. Both methods use OOP health spending as the numerator. Typically, OOP excludes third-party reimbursements and insurance payments. The denominator in the budget-share method is household income or consumption, while it is CTP in the WHO CTP method. CTP is the amount left for a household to spend after basic (or subsistence) needs are met. For a household, (i), with total consumption expenditure ({ THC}_{exp }), its CTP is expressed as:

where ({SE}_{i}) stands for expenditure on subsistence (basic) needs, which is the minimum requirement to maintain basic life in a society. ({SE}_{i}) is calculated from a defined basic need line, ({Pl}_{n}) and the equivalent household size, ({eqsize}_{i})(which adjusts for household size and composition according to the economy of scale in consumption needs), using the equation:

The original WHO CTP method, known as the WHO standard method, is the most frequently used method in studies in Nigeria [34]. The WHO European regional office recently pioneered an improved CTP method called the WHO-Europe method [10]. The WHO standard method is partially normative because it recognises only food as a basic need. The WHO-Europe method is fully normative, including other basic needs such as housing and utilities (water, electricity, cooking gas and other fuels for heating) [10]. Conventionally, restaurant food, tobacco, and alcohol are not considered basic needs [16].

The WHO-Europe method differs from the WHO standard method in three main ways. First, it uses a poverty line which reflects spending on fully normative basic needs among relatively poor households (those between the 25th and 35th percentiles of the household consumption distribution) in a country [10]. The WHO standard method, however, uses food expenditure among households in the average consumption distribution (the 45th and 55th percentiles) [15, 35]. Second, unlike the WHO standard method, the WHO-Europe method allows very poor households with total household consumption expenditure below the subsistence line to have negative CTP [10]. Instead, the latter substitutes actual food expenditure for the higher SE [16]. Third, the WHO-Europe method uses the OECD equivalence scale [10, 36] instead of the WHO standard method’s scale [15].

The WHO Europe method provides the benefit of using a basic need line closest to national poverty lines [10]. Moreover, it is more equitable than the budget-share and WHO standard methods, which are pro-rich, reducing CHE among the poor [17]. The budget share method, adopted as the official FP monitoring indicator for SDG 3.8.2, is worse off in this respect [10, 35]. WHO Europe method’s proponents assented to its broader applicability in high- and middle-income countries [10]. Studies from Nigeria and other LMICs, such as India and Bangladesh that used similar methods attest to this fact [21, 37,38,39].

Given the weaknesses of the other methods, we leveraged the WHO-Europe method to produce actionable evidence of financial hardship from NCDs in Nigeria [10, 17]. We adapted this method for Nigeria in two ways. First, electricity and heating fuel were removed from the list of basic utilities, leaving cooking fuel. House heating is not a basic need in Nigeria, and only 59% of households have electricity [40]. Second, clothing was introduced because it was adjudged a necessity [37]. As recommended in the original WHO-Europe method [10, 17], we only included actual rent in our calculations.

The WHO-Europe and WHO standard thresholds are 40%, and the budget share threshold is 10% [41]. Thresholds are arbitrary. Rashidian et al. (2018) [42], using the Receiver Operator Characteristic (ROC) curve and Kappa method, computed 20% and 35% as the optimal thresholds for the budget share and CTP methods, respectively. Using different thresholds by researchers produces very different rankings among countries, impeding global monitoring [41]. For sensitivity purposes, following Rashidian et al. (2018) [42], we varied the thresholds used: 20%, 25%, 30%, 35%, and 40%. For comparison with other studies, we have included CHE estimates at different thresholds for the budget share and WHO standard methods (Supplementary Table 2).

Estimation of household impoverishment

Household impoverishing health expenditure occurs if a household is impoverished by OOP health spending. Households are classified into five categories based on their remaining consumption expenditure relative to the poverty line after OOP health expenses [10, 35]. We used Nigeria’s 2018 national poverty line of ₦137,430 (US$447.78) [43]. A household is “impoverished” when its total consumption, though above the poverty line before OOP health spending, falls below the poverty line after OOP spending. A household whose total consumption was already below the poverty line before incurring OOP health payments is said to be “further impoverished”. “At-risk of impoverishment” and “not-at-risk of impoverishment” households have total consumption above the poverty line before or after OOP health spending. A household’s total consumption relative to 120% of the poverty line after OOP health payments determines whether it is “at risk of impoverishment” or not. Households whose total consumption falls on or below this 120%-line (but above the poverty line) are “at risk of impoverishment”, and those who remain above it after OOP spending are “not at risk of impoverishment”. The “non spender” group had no OOP spending. They could fall above or below the poverty line. This group includes, among others, those who missed care due to cost [35].

WHO-Europe method highlights three important categories of households obscured in the traditional impoverishment measurements [44]. These groups include the “further impoverished”, “at risk of impoverishment”, and “non-spenders”. The first two are vulnerable groups as crucial as the impoverished category []. Not considering the “further impoverished” suggests OOP health expenditures are only harmful if they cause poverty, not if they worsen it. Also, the “at risk of impoverishment” group is critical in poverty-prevention policies [10]. A high percentage of the “non-spender” group indicates significant forgone care in a population [35]. In this case, low CHE estimates could give a false impression that the population enjoys FP [10]. Forgone care worsens health conditions, lowering the household’s productivity and welfare [12].

Estimation of productivity loss

Household productivity loss was estimated using the methods suggested in the WHO Global TB Programme protocol [18]. We used an input-based human capital approach to calculate productivity loss for a household, (i) with (n) members:

Where, ({P}_{loss}) is the total annual productivity loss of an individual in the household, ({t}_{confinement}) is the number of days the individual is confined to a hospital bed, ({t}_{missed}) is the number of days the individual missed from primary activities, (apart from days confined to a hospital bed), ({w}_{total}) is the daily income of an individual and the sum of the time, ({(t}_{confinement}+{t}_{missed})) is the total work-time lost.

The number of days individuals missed from their usual activity and the number of days confined to bed had recall periods of 30 days and 12 months, respectively, in the survey. The former was annualised by multiplying by 12.17 [33].

Our productivity loss estimation is from ill health, not death and relates to personal but not employer-related losses. We assumed there was constant income, no payment for sick-off periods, no social safety nets or health insurance and no compensatory input from other household members. Nigeria has less than 4% health insurance coverage [42] and little social protection against illness [45]. Also, we assumed an informal caregiver’s income loss is equal to the patient’s. This may not be true for informal and seasonal workers.

The debate in the human capital approach for productivity loss calculation has been about the appropriateness of income used [46, 47]. In this study, we employed three approaches proposed by the WHO [44] to determine our income values. First, respondents’ pre-illness incomes were used. We imputed for missing data the average income of people in their income quintile. This approach has, however, been blamed for equity issues [18, 46, 47]. Multiplying the unadjusted productivity loss by the sample’s labour force participation rate, 72.21% (95% CI: 71.85–72.58), corrected for unemployment. Second, we used the lowest-paid unskilled government worker pay, $30,000 (US$9775), in 2018 [48]. Third, we used Nigeria’s annual GDP per capita of ₦ 622,372.38 (US$2027.8) in 2018 [49] instead of the minimum wage [47]. We accounted for unemployment in the last two methods using half the general wage as suggested by WHO (2015) [18]. The three methods jointly reduce errors arising from imputed income, improving the spread and sensitivity of our estimates.

Outliers (about 0.05%) in our dataset were identified and removed before all analyses, using the trimming method. We estimated the OOP expenditure, CHE and impoverishment at the household level. The sampling weight supplied with the data was used in all our analyses for national representativeness. Data analysis was performed using Stata MP, version 16.0 (StataCorp, Texas, USA).