Effects and determinants of tuberculosis drug stockouts in South Africa

Medicine shortages affect countries worldwide, but in particular developing countries [1‐4] The evidence to date is limited but suggests that stockouts are associated with treatment alteration and discontinuation, and with worse health outcomes, but none of these studies provides any estimates of the impact of stockouts on health outcomes. It also does not examine whether stockouts affect rich and poor communities differentially.

The drug stockout problem is particularly pressing in South Africa for TB medicines for four reasons. First, South Africa is facing a huge TB epidemic, with 438,000 cases a year and an annual mortality of 123,000 people [5]. Secondly, South Africa has very high levels of drug stockouts with 14% of the primary health care facilities reporting a TB drug stockout in 2011 [6]. Thirdly, TB is a lethal, but treatable disease and drugs are dispensed for free to TB patients by the public health clinics, but periods of stockouts during the six-month treatment period could have a severe adverse impact on treatment success. Finally, the country has notoriously high levels of income inequality [7] which makes it likely that the drug stockouts do not affect richer and poorer districts equally.

We use two data sources that were linked at the district level. The District Health Barometer [24] was used to extract TB drug stockout data, TB health outcomes and health system characteristics. This dataset contains annually aggregated indicators of public primary health facilities in South Africa for the years 2011, 2012 and 2013 at the district level. Socioeconomic information was obtained from the Census survey 2011 [25] household data and then aggregated at the district level. Variable definitions and descriptive statistics can be found in Tables 1 and 2. Of the 52 districts in South Africa 51 were included, as stockout proportions were missing for Tshwane District. Stockout proportions were also missing for Eden and Overberg in 2012 and for Central Karoo and Cape Town in 2011 (all located in the Western Cape Province).

There is substantial variation in TB stockout proportions across provinces and across the years 2011, 2012 and 2013 (Fig. 1). In our study period, the Western Cape was on average the province with the lowest and the Northern Cape the province with the highest stockout proportions (3.84 vs. 16.52).

Table 3 lists the average TB stockout proportions by district and TB treatment outcomes by district by poverty categories. TB stockout proportions are higher and TB treatment outcomes are seen to be worse in districts with higher poverty. The ANOVA confirms that TB outcomes vary by level of poverty of the district.

Our study is the first that has tried to exploit the variation in drug stockouts across districts and over time in South Africa to examine the determinants and the effects of TB drug stockouts.

First, we found that TB drug stockouts are more prevalent in districts with lower income: a 10% rise in income was associated with an 8.5% decline in TB drug stockouts. This suggests that poorer districts are likely to suffer a greater burden of stockouts.

Secondly, and contrary to a priori expectation, we found that districts in which the Provincial Medicine Depot was located were more likely to experience a TB drug stockout. This suggests that greater distance to a Provincial Medicine Depot does not appear to operate as a constraint on adequate stock levels in South Africa. It also suggests that facilities located in the same district as the provincial depot overly reliant on the provincial depot.

Third, after controlling for time invariant unobservables, we estimate that TB stockouts had substantial negative effects on TB cure and success rates by district. Our results indicate that, on average, a 10% rise in the TB drug stockout proportion results in a 2.14% decline in TB cure rate and a 1.43% decline in TB success rate. These negative consequences of TB drug stockouts were larger in the districts with higher poverty rates. This confirms that the negative effects are primarily concentrated in the poorest districts of South Africa and are likely to further exacerbate the socioeconomic inequalities in TB treatment outcomes.

Our results imply that TB drug stockouts can limit the effective control of the TB epidemic, especially in districts with higher rates of poverty. This is because, when fewer patients are cured, more patients remain infectious and are able to spread TB. Improved reliability of the TB drug supply is therefore essential to enhance the effectiveness of TB epidemic control.

The finding that TB drug stockouts seem to have a greater negative impact on TB cure rates than on TB success rates deserves further discussion. The TB cure rate measures the fraction of all smear-positive TB patients who are smear negative in the last two months of their treatment. The TB success rate measures the fraction of all TB patients who completed their treatment and all smear-positive TB patients who are smear negative in the last two months of their treatment.

The South African public health system transitioned to Genexpert⁴ in 2012 which means that a small minority of patients receive smear tests. For patients who did not receive smear tests the success rate merely captures whether they have had a recent clinic visit to collect their treatment. Patients included in the TB success rate are therefore not necessarily cured. For TB to be cured, the treatment regimen must be followed strictly and interruptions can have serious consequences for cure [21‐23], whilst for TB success treatment must be completed. Interruptions may therefore have less impact, as reflected in our results.

The finding of a negative impact of TB drug stockouts and TB death rates in the poorest districts is counterintuitive, but the further analysis with lost to follow up patients suggests that the initial finding may be attributable to missed deaths and the higher concentration of such missed deaths in poorer districts.

We found that the effects of TB drug stockouts are worse in districts with more people living in poverty. This could be due to the fact that in these districts patients have fewer options to look for alternatives to access TB drugs. When a health facility has a stockout, patients can go to another health facility to collect their TB drugs. This requires extra travel and involves time and money costs. Another option would be to buy the TB medicine at a private pharmacy or clinic. When these options are not feasible for cash-constrained patients, they are forced to interrupt TB treatment until the TB drugs are available.

TB medicines were not the only drugs that have had stockout problems in South Africa. Stockout surveys [29‐31] have shown that many other essential medicines were also affected. This problem was acknowledged by the National Department of Health and some measures have been taken to curb the problem [32]. However, and in spite of these, 20% of health facilities in South Africa still reported stockouts of essential medicines (including TB drugs) in 2016 [33].

Since the consequences of TB drug stockouts for TB treatment outcomes are severe, emphasis should be placed on further strengthening of the pharmaceutical chain in South Africa. A short-term solution to ensure continuous drug supply to TB patients could be a public-private partnership in which TB patients can collect their TB drugs for free at a private pharmacy or -clinic in the case of a stockout in the public health facility.⁵

While contributing to the literature on the subject, our study also has limitations. First, the aggregated annual stockout proportion at the district level is measured with substantial imprecision. The district stockout proportion was not weighted by clinic size and does not include any information about the duration of the stockout. As a result, a stockout of a few days is weighted equally to a stockout of a month. Since the publicly available data were already aggregated and not available at the facility level, it was not possible to adjust and refine these rates. Secondly, as our analysis was performed at the district level, it may suffer from ecological fallacy or bias. Some districts are quite large and considerable variation in TB drug stockout proportion and TB treatment outcomes between clinics within districts exist. TB drug stockouts can occur in one clinic, whereas bad TB treatment outcomes can occur in another clinic. These differences disappear when TB drug stockout proportion and TB treatment outcomes are averaged on a district level. A potential future analysis at facility level could solve this problem but will run into another problem: that of attributing TB patients to clinics. Last but not least, the fixed-effect regression analysis controlled for time invariant unobservables, TB incidence and HIV coinfection. To the extent that any other relevant time varying unobservables were omitted, our estimates may still be subject to some omitted variable bias.

The persistence of TB drug stockouts in South Africa not only hampers the effective control of TB but its unequal occurrence across districts also adds to the socioeconomic inequality in TB outcomes. We find that stockouts are not only more prevalent in poorer parts of South Africa, they also have greater detrimental effects on the TB treatment outcomes of the poorer districts. This suggests that efforts to reduce TB drug stockouts are essential in the fight against the TB burden and its impact on socioeconomic inequality.