Increasing Prescription Length Could Cut Cardiovascular Disease Burden And Produce Savings In South Africa

Abstract

South Africa's rates of statin use are among the world's lowest, despite statins’ demonstrated effectiveness for people with a high blood cholesterol level or history of cardiovascular disease. Almost 5 percent of the country's total mortality has been attributed to high cholesterol levels, fueled in part by low levels of statin adherence. Drawing upon experience elsewhere, we used a microsimulation model of cardiovascular disease to investigate the health and economic impacts of increasing prescription length from the standard thirty days to either sixty or ninety days, for South African adults on a stable statin regimen. Increasing prescription length to sixty or ninety days could save 1,694 or 2,553 lives per million adults, respectively. In addition, annual per patient costs related to cardiovascular disease would decrease by $152.41 and $210.29, respectively. Savings would largely accrue to patients in the form of time savings and reduced transportation costs, as a result of less frequent trips to the pharmacy. Increasing statin prescription length would both save resources and improve health outcomes in South Africa.

Cardiovascular disease is the leading cause of death globally. Even in African nations where burdens due to communicable diseases such as HIV and tuberculosis are prominent, cardiovascular disease is the second leading cause of death, and it is the leading cause of death for adults over the age of fifty.¹

High cholesterol, a risk factor for cardiovascular disease, is poorly controlled in low- and middle-income countries. In 2000 in South Africa high cholesterol caused 59 percent of ischemic heart disease (angina, myocardial infarction, and cardiac arrest) and 29 percent of ischemic stroke in adults over the age of thirty.² In addition, 4.6 percent of the total mortality in South Africa was attributable to high cholesterol.² The burden is high because widespread screening for cardiovascular disease is inadequate, levels of initiation of treatment for those eligible are low, and so are levels of adherence once treat ment is initiated.^3,4 By adherence or persistence, we refer to the degree of conformity to the recommendations about day-to-day treatment by the provider and the frequency of medications taken.

Statins are the most commonly used and most cost-effective drugs for high cholesterol.⁵ Although statin use is generally low worldwide, South Africa has extremely low initiation and utilization rates compared to those of other countries.^3,4 A study whose subjects were recruited from 2003 to 2005 found that South Africa had the lowest rate of statin use (55.1 percent) in secondary prevention at thirty days after discharge from the hospital following a cardiovascular disease event, compared to both high-income countries and low- and middle-income countries (range: 68.8–89.9 percent).³ Secondary prevention in this context refers to the use of medication after a person survives a cardiovascular disease event.

Some of the low usage rates in South Africa are due to low levels of initiation of medications in the first place. However, even among those who receive and fill prescriptions and are initially adherent, many do not remain adherent. In 2011, 48 percent of the patients in South Africa who received statin treatment failed to reach the recommended lipid levels, in some cases because of poor adherence.^6,7

A qualitative study provides information on why South Africans stop taking their medications.⁸ Some of the factors identified in the study were distance to health care facilities, lack of transportation, provider prescribing patterns, and financial challenges (such as the cost of transportation and medication and loss of income because of a clinic visit). In many lowand middle-income countries, patients must walk long distances to government-run pharmacies, spend hours in line (averaging up to four hours),⁹ and miss significant time at work.

Interventions to address these problems could lead to improvements in adherence rates. A study in Denver, Colorado, presented a potential intervention that increased the quantity of statin dispensed at each prescription refill.¹⁰ Patients who were given longer prescription lengths of statin at each refill had better adherence rates and treatment effectiveness, compared to patients with shorter prescription lengths.

Prescription length in South Africa is typically set at 28–31 days. Increasing this period to two or three months for patients who have been on a consistent regimen for at least six months could reduce the number of visits to pharmacies, while improving statin adherence and treatment outcomes.¹¹ In this analysis we used a microsimulation model of cardiovascular disease to assess the economic and health impacts of increasing prescription length for people on a stable statin regimen.

Study Data And Methods

INTERVENTION

The intervention that we modeled was to provide South African patients taking sta-tins with prescriptions for longer time periods than the current norm. We assumed that for patients to be eligible for increased prescription length, they would first need to be on a stable statin regimen for six months. This could reduce the potential for wastage by reducing the likelihood of change in medication type or dose.

To assess the effects of increasing prescription length from the standard monthly prescription in South Africa, we compared the baseline thirty-day prescription length with sixty- and ninety-day prescription lengths.¹²

INTERVENTION EFFECTIVENESS

In a review of the Denver study, we found that increasing prescription length from thirty to sixty days led to approximately a 9 percent increase in adherence as well as a 6 mg/dL reduction in cholesterol level.¹⁰ Using data from the province of Ontario¹³ and the state of California,¹⁴ we found that increasing the length from sixty to ninety days produced 4.5 percent more adherence and a reduction of 3 mg/dL in cholesterol level.

These reductions in cholesterol are estimated to lead to reductions of approximately 1–5 percent in the relative risk for the various cardiovascular disease event rates.¹⁵ Cardiovascular disease events are stroke, myocardial infarction, and death. The relative risk reductions for each prescription length and the ranges we used for our multiple sensitivity analyses are reported for each event type in Exhibit 1. Further details on the reductions in cardiovascular disease from increased adherence are available in the online Appendix.¹⁶

EXHIBIT 1.

Relative Risk (RR) Reduction Of Cardiovascular Disease, By Statin Prescription Length

	Prescription length (days)
	30		60		90
	RR reduction	Reported 95% CI	RR reduction	Sensitivity analysis	RR reduction	Sensitivity analysis
primary prevention
MI	0.77	(0.74, 0.80)	0.73	(0.72, 0.75)	0.71	(0.70, 0.74)
CVA	0.83	(0.78, 0.88)	0.80	(0.80, 0.81)	0.79	(0.80, 0.81)
secondary prevention
MI	0.76	(0.73, 0.79)	0.72	(0.71, 0.74)	0.70	(0.70, 0.73)
CVA	0.85	(0.81, 0.90)	0.82	(0.82, 0.83)	0.81	(0.80, 0.83)
Death	0.85	(0.77, 0.95)	0.83	(0.83, 0.84)	0.82	(0.81, 0.83)

source Authors' analysis of data from the following sources: Baigent C et al. Efficacy and safety of cholesterol-lowering treatment (Note 30 in text); Mihaylova B et al. The effects of lowering LDL cholesterol with statin therapy in people at low risk of vascular disease (Note 31 in text); and Batal HA et al. Impact of prescription size on statin adherence and cholesterol levels (Note 10 in text). notes Primary prevention is the prevention of onset of symptomatic disease. Secondary prevention is the use of medication after a patient has survived a cardiovascular disease event. CI is confidence interval. MI is myocardial infarction. CVA is cerebrovascular accident (stroke).

Simvastatin was the primary statin of choice in our model since it is the medication recommended for initial treatment in the essential drug formulary for the South African public sector. However, many patients on protease inhibitors or other medications are switched to atorvastatin because of interactions between protease inhibitors and simvastatin. Thus, we also conducted sensitivity analyses using atorvastatin, which is a second-line agent on the South African essential drugs list. Side effects of the statins—which include myalgia (4.7 percent), rhabdomyolysis (0.006 percent), and the risk of developing diabetes (odds ratio: 1.09)—and their rates were included in the model.

MODEL POPULATION

Our model was populated with data on 1,214 patients from a pilot study that evaluated community health workers’ effectiveness in using a noninvasive cardiovascular disease risk tool, compared to the effectiveness of physicians’ and nurses’ use of the same tool.¹⁷ The patients (ages thirty-five and older) came from underserved communities in the townships surrounding Cape Town, South Africa. Patients in the model were without a history of a previous cardiovascular event (for example, stroke or myocardial infarction) and without a history of a previous diagnosis of diabetes or hypertension. Patient characteristics are presented in Table 1 in the Appendix¹⁶ and are similar to those in a nationally representative sample.^18,19

MODEL DESCRIPTION

We used a cardiovascular disease microsimulation model that was initially developed and validated for the United States, and we calibrated the input parameters to adjust it for the South African population.^20–22 That population is briefly described here; for additional details, see the Appendix.¹⁶

Our model population was expanded from 1,214 patients to a million representative individuals and simulated year by year until death. South African all-cause mortality, cardiovascular disease–specific mortality, and each individual's cardiovascular disease risk factors (age, sex, systolic blood pressure, body mass index, smoking status, and diabetes status) determined annual probabilities of death and cardiovascular disease events.²³

Cardiovascular disease event probabilities were generated using previously published cardiovascular disease risk equations for both the United States and South Africa.^24–26 The cardiovascular disease health states simulated in the model were coronary heart disease and stroke or cerebrovascular accident. Coronary heart disease states were further divided into myocardial infarction, angina, and resuscitated cardiac arrest.

The model also incorporated screening and treatment. During screening, an individual's cardiovascular disease risk was estimated. For those with more than a 15 percent risk of a cardiovascular disease event over five years, statins were recommended as primary prevention, according to the South African dyslipidemia guidelines.²⁷ Medication reduces an individual's probability of experiencing a cardiovascular disease event. Secondary prevention was modeled similarly.

INTERVENTION COSTS

The direct costs of the intervention are reported in Exhibit 2. Patient costs were time costs and travel costs incurred for traveling to and from the pharmacy.

EXHIBIT 2.

Annual Statin-Associated Costs Per Person, By Prescription Length

	Prescription length (days)
Type of cost	30	60	90
Transportation	$12.59	$ 7.06	$ 4.96
Time cost
Travel	16.38	9.18	6.46
Pharmacy	61.66	34.59	24.31
Labor cost
Pharmacist	15.47	8.68	6.10
Medication cost
Simvastatin (20 mg)	6.65	7.46	7.87
Atorvastatin (20 mg)	22.16	24.86	26.21

source Authors’ analysis. note All costs were adjusted to 2012 US dollars.

We multiplied the annual expected number of pharmacy visits by adherence rates to calculate the actual number of visits to the pharmacy. For example, instead of assuming that a person with a thirty-day prescription visited the pharmacy twelve times per year, we multiplied the expected number of pharmacy visits by the adherence rate with a thirty-day prescription length, or 73.9 percent, which resulted in 8.868 pharmacy visits per year. Similarly, on average, a person in the sixty-day refill group, with an 82.9 percent adherence rate, would make 4.974 visits a year; and a person in the ninety-day refill group, with an 87.4 percent adherence rate, would make 3.496 visits.

Health system costs either induced or offset by the intervention were outpatient costs and the costs of hospitalizations, procedures, and caring for people with chronic conditions. Details for the bottom-up costing of the remaining inputs are described in the section of the Appendix on the cost of the intervention.¹⁶

All costs were adjusted to 2012 US dollars.We used South African Consumer Price Index values²⁸ and exchange rates from World Bank²⁹ to make the adjustments.

We also estimated the cost of wastage. Medication wastage can result from changes in prescribed medication, poor adherence, or death.¹⁴ In the California study mentioned above,¹⁴ the additional wastage from ninety-day prescriptions compared to thirty-day ones was estimated at approximately three days of medication per person per year. We incorporated this loss into the medication costs. Given some estimates that wastage could be as high as 10 percent,³⁰ we incorporated these additional costs at the higher end into a sensitivity analysis.

SENSITIVITY ANALYSIS

We did multiple sensitivity analyses on the key variables. First, we conducted a sensitivity analysis on the effectiveness of the intervention—namely, the impact of increasing prescription lengths on adherence rate and, ultimately, on cholesterol levels. We tested whether the intervention would be 33 percent more or less effective than results seen in North America. Specifically, we tested both the benefit that we estimated to be a 6 mg/dL cholesterol reduction across the range from a 4 mg/dL reduction to a 8 mg/dL reduction for the change from thirty- to sixty-day prescriptions and the estimated benefit of an additional 3 mg/dL reduction across the range from a 2 mg/dL reduction to a 4 mg/dL reduction, with their associated relative risk reductions (Exhibit 1).

Second, we did a sensitivity analysis on the cost of the intervention, ranging from 50 percent below to 50 percent above the cost in the base case. Third, we assessed the effectiveness of the statins across the 95 percent confidence interval of the relative risk reduction of each statin.^15,31

Fourth, in another sensitivity analysis, the wait time at pharmacies was set at a lower threshold of thirty minutes, down from four hours. Lastly, given the documented interaction between simvastatin and protease inhibitors, we conducted a sensitivity analysis in which we replaced simvastatin with atorvastatin in anticipation of increased use of the latter in the future.³²

LIMITATIONS

There are some limitations to our analysis. The benefit from increasing prescription length in terms of improved adherence and reduced cholesterol levels is based on estimates from studies in North America.^10,13,14 However, we have data showing that the study population's overall adherence rates are similar to those of US populations with chronic conditions. Subjects in the Denver study had an adherence rate of 73.9 percent (using the medication possession ratio method, adherence was defined as being in possession of 80 percent of the total prescribed medications).¹⁰ A study conducted in Soweto has shown that 71 percent of patients with chronic heart failure have been adherent to their regimen. The method the researchers used to determine adherence was counting prescribed pills: If 75 percent or more of the prescribed pills were taken, the patients were defined as being adherent.³³ Also, all three North American populations were underserved and received their care in the public sector, similar to the population we evaluated in South Africa.

Furthermore, increased prescription length might be inappropriate for patients who prefer or need to be seen more frequently when initiating statins or followed closely for medication effects.³⁴ However, to minimize the potential cost associated with an increase in the number of visits after an initial prescription of statins, we modeled the increased prescription lengths only for people who had been on a stable dose for at least six months. In addition, we conducted a sensitivity analysis on intervention effectiveness to address the concern that the North American data might not be completely generalizable to the South African population.

There were costs that were not considered in our analysis. We did not include the cost of educating health personnel about the new policy and the government cost of changing a policy. This was because we had insufficient information on how health personnel are educated about new regulations and because of the lack of standardized methods to quantify government costs incurred when implementing a new regulation. As a result, we assumed that those costs are negligible, which means that we might have underestimated the cost of increasing prescription lengths.

Also, we did not include the medication-dispensing fee in our analysis because of limited data about its amount. However, we know that the dispensing fee is much higher with thirty-day prescriptions than with longer prescriptions. For example, the fee occurs nearly three times as frequently with thirty-day prescriptions as it does with ninety-day prescriptions. If we had considered the dispensing fee, the cost savings could have been bigger. Several studies have shown that reductions in the dispensing fee far outweigh any wastage loss, particularly with generic medications.^14,30,35

We also did not include the potential impact on other medications if the policy of increased prescription lengths were systemwide and also included medications for conditions such as hypertension and diabetes. If making the policy systemwide produced the same adherence effects with increased benefits for other conditions, then we could have significantly underestimated the effect of such a policy change.

Study Results

Our results show that significant health benefits can be gained by increasing adherence through increased prescription length. An increase from thirty days (the standard) to sixty days can avert over 1,000 ischemic heart disease deaths and over 650 stroke deaths per million population (Exhibit 3). In addition, the change can avert an additional 3,000 nonfatal myocardial infarctions and strokes per million population. If the prescription length were ninety days instead of thirty, even more deaths and nonfatal cardiovascular disease events could be averted.

EXHIBIT 3.

Fatal And Nonfatal Lifetime Events Averted Per Million Adult Population, Compared To 30-Day Prescription Length

	Prescription length (days)
Number averted	60	90
MI events	2,039	3,045
CVA events	1,049	1,556
IHD deaths	1,016	1,536
CVA deaths	678	1,017

source Authors’ analysis. notes MI is myocardial infarction. CVA is cerebrovascular accident (stroke). IHD is ischemic heart disease.

The total cost savings to society are $150 per capita over the course of a lifetime for replacing a thirty-day with a sixty-day prescription length and over $200 for switching to a ninety-day prescription length (Exhibit 4). Therefore, changing to the sixty- or ninety-day prescription length could save, on average, $6 and $9 per capita per year, respectively, in health care costs and patients’ direct time costs combined. The ninety-day prescription length saved the most of all three strategies and had the greatest impact on increasing quality-adjusted life-years, compared to current standards.

EXHIBIT 4.

Components Of Total Cost Savings Compared To 30-Day Prescription Length

	Prescription length (days)
	60		90
Type of cost	Cost savings	Percentage	Cost savings	Percentage
Transportation	$ 18.27	12.0	$ 25.25	12.0
Patient direct time	113.24	74.3	156.45	74.4
Health system	20.90	13.7	28.59	13.6
Labor (pharmacist)	22.43	– a	31.01	– a
Medication	−2.90	– a	−4.27	– a
CVD event	1.37	– a	1.85	– a
Total	152.41	100.0	210.29	100.0

source Authors’ analysis. notes All costs were adjusted to 2012 US dollars. Cost of medication includes wastage cost. A negative cost saved means that the cost of a longer prescription was more than the cost of a thirty-day prescription. CVD is cardiovascular disease.

^aNot applicable.

The majority of the cost savings in the intervention came from reducing transportation costs and patients’ direct time costs as a result of fewer pharmacy visits. Furthermore, the sixty-and ninety-day prescription lengths led to even greater reductions in overall expenditures stemming from the reductions in ischemic heart disease and cerebrovascular events. Approximately 74 percent of the savings were due to patient direct time savings, 14 percent to reductions in health system spending at the pharmacy or hospital because of reduced events, and 12 percent to reductions in transportation spending (Exhibit 4).

Wastage resulted in a reduction in the savings by only $0.25, assuming 1 percent wastage for both sixty- and ninety-day prescription lengths compared to thirty-day lengths. When we assumed an outer limit of 10 percent wastage, the savings were reduced by slightly less than $2.55. In neither case did wastage change the rankings of the three prescription lengths or significantly affect the overall savings.

When we changed the intervention effectiveness by one-third in either direction in a sensitivity analysis, as described above, we found no significant changes in the rankings of the three prescription lengths or the overall results. The lower bound of the reduction in cholesterol level saved $151.77 in per capita lifetime costs for the sixty-day prescription length and $209.23 for the ninety-day prescription length, compared to the thirty-day prescription length. The upper bound saved $152.79 and $210.96, respectively.

When we changed the cost of the intervention in another sensitivity analysis, also described above, the rankings of the interventions did not change, but the magnitude of the benefit did. Compared to the thirty-day prescription length, the cost savings ranged from about $78 to $228 per capita for the sixty-day prescription length and from about $107 to $314 for the ninety-day prescription length.

The sensitivity analysis on the effectiveness of the statins also did not change the strategies’ rankings. The sixty-day prescription length had cost savings ranging from $151.43 to $153.71 per capita, and the ninety-day prescription length had cost savings ranging from $208.97 to $212.15, compared to the thirty-day prescription length.

Average wait times have been estimated to be four hours in the base case. When we did a sensitivity analysis with a thirty-minute wait time, the sixty- and ninety-day prescription lengths were preferred to the thirty-day prescription length. However, the savings per capita were about half those with the four-hour wait time, or about $76 and $105 per capita, respectively, over an individual's lifetime. Finally, when we substituted atorvastatin for simvastatin, the sixty- and ninety-day prescription lengths were preferred to the base case, with $146 and $201 saved, respectively.

Discussion

Our study aimed to analyze the economic and health outcomes of increasing the prescription length from thirty days to sixty or ninety days to improve statin adherence rates in South Africa. The results of this study indicate that increasing prescription length would not only be more effective in reducing cardiovascular disease burden but would also be cost saving in South Africa.

The health gains from the longer prescription length were measured in improvements in the relative risk reduction (lower values are better). The changes were small in absolute terms, with a 2–6 percent reduction in relative risk of cardiovascular disease event rates from the thirty-day prescription length rates to either the sixty- or ninety-day rates. However, the number of stroke and ischemic heart disease events is large. If the prescription length of ninety days were applied (after at least six months of stable use) to all adult South Africans who were eligible for sta-tins, up to 90,000 cardiovascular disease deaths could be averted over the adults’ lifetimes.

In addition to the health benefits of the intervention, significant savings could be realized by individuals, the health care system, and society. The majority of the savings would go to the pa tient through decreased transportation expenditures and reduced time costs. Even in a country with high unemployment rates, the cost savings were substantial. For employed patients, the four-hour wait times at pharmacies in the public sector⁹ necessitate missing nearly an entire day of work, which would happen once a quarter instead of once a month if prescription lengths were changed from thirty to ninety days. Pharmacy wait times could be reduced with new technologies or more efficient scheduling, but there are still inefficiencies with monthly visits, in terms of travel time lost and transportation cost. Furthermore, the health system would enjoy reductions in expenditures from increased prescription length through reduced pharmacist time and reduced medical care as a result of a decrease in cardiovascular disease events.

Our study was a modeling study, but an analysis of a retrospective cohort in the United States showed similar results.³⁶ In that case, the authors found that in a population of over 130,000 patients, those with over 80 percent adherence to medications for hypercholesterolemia had less than half of the medical costs of patients with less than 20 percent adherence, with an inverse linear trend across the groups. Some of the cost savings could be less significant if patients returned to the clinic more frequently in the first few months after a cardiovascular disease event or because a pharmacist may be able to help with medication complications in the first six months. However, limiting increased prescription length to patients who have stabilized on a current dose for at least six months would minimize the likelihood of these two scenarios or of wastage having an a large impact.

There are many barriers to medication adherence, some under the patient's control and others under the control of the provider or the health system.³⁷ However, prescription length appears to be even a stronger barrier than co-payments.

One study evaluating the impact of two policy changes in the Medicaid program in North Carolina showed that when the state decreased prescription length from a hundred to thirty-four days and introduced copayments for brand-name medications, both changes led to decreased adherence to medications for chronic conditions.³⁵ However, the adherence reduction in four classes of medications, including statins, showed that the effect of the change in length of prescription was greater than the effect of the introduction of copays. This indicates that the impact of increased costs (time and transportation) associated with more frequent trips to the pharmacy was greater than the impact of the introduction of copays.

Increased prescription length is popular among both insurers and patients. In the first year after Walmart introduced the option of a ninety-day length, national rates of ninety-day prescriptions increased by 60 percent for both Medicare and commercially insured patients.³⁸

Wastage has been mentioned as another concern when prescription length is increased. However, the evaluation of the change made in the California Medicaid program showed only 1 percent more wastage when prescription length was increased.¹⁴ Even when we tested the extreme case of estimated wastage of 10 percent more in our study, increased prescription length was still a cost-saving intervention, and overall savings were reduced by only 1 percent.

Two considerations explain the relatively minimal impact of wastage today. First, the overall cost of generic medications today is much lower than that of patented medications ten years ago, which blunts the impact of wastage. For example, statins in South Africa cost $0.03–$0.08 per patient per day. Second, in studies where wastage was reported, the associated cost was dwarfed by savings in dispensing fees that were also reported, because of the lower numbers of refills and pharmacy staff needed.¹⁴ In the United Kingdom increasing dispensing frequency to every month from every three months would increase the pharmacy budget by one-third.³⁰

One strength of our study is that it used a relatively easy intervention that saves resources and improves health at the same time. Most of the interventions that address medication adherence problems (such as patient education and monitoring) require additional time and health system expenditure.³⁹

In contrast, increasing prescription lengths also reduces overall patient and health system costs and is more convenient to both patients and health personnel. The intervention also reduces patients’ frequency of travel, waiting times, and loss of workdays. Furthermore, it saves pharmacists time, which leads to decreased wait times overall.

Another strength of our study is that to our knowledge, it is the first cost-effectiveness study on managing cardiovascular disease risk levels by improving medication adherence through an increase in prescription length. Because it was conducted in South Africa, our study shows how governments in resource-scarce settings could improve cholesterol management.

Conclusion

In the management of chronic conditions, there is no single magic bullet. More likely, a series of small reductions in cardiovascular disease events will be needed if the world is going to meet the overall World Health Assembly goal of a reduction of noncommunicable diseases by 25 percent by 2025.⁴⁰ Smoking cessation will be important. Efforts to increase screening will also be important as a way to increase awareness among those with chronic conditions who might benefit from intervention. Next, providers will need to be encouraged to initiate the use of medications such as statins for people who would benefit from them, in terms of both primary and secondary prevention. Efforts to improve adherence in ways that include increasing prescription length would help set in motion this cascade of events that is necessary to decrease the burden of cardiovascular disease.