“SHAKE-SHAKE IT OFF”: IMPLEMENTING SBIRT IN MWEMBESHI AND CENTRAL PROVINCE, ZAMBIA Prepared for the Lutheran Health Alliance By: Kate Austin Michelle Duren Ben Emmel Diana Rosales Alex Straka Public Affairs 881: Benefit-Cost Analysis December 19, 2014 ACKNOWLEDGEMENTS We would like to thank Dr. Jason Paltzer for his assistance and persistent resourcefulness throughout the project. Additionally, we would like to thank both Kingdom Workers and the Lutheran Health Alliance for their cost-benefit analysis request. The project provided us with a wealth of knowledge and an opportunity to apply our cost-benefit analysis skills to a complex and engaging real-world scenario. Finally, we would like to thank Professor David Weimer for his counsel, support, and direction throughout this project. ii TABLE OF CONTENTS Acknowledgements ................................................................................................................ ii Table of Contents .................................................................................................................... iii List of Tables ............................................................................................................................ iv Executive Summary ............................................................................................................... vi Introduction .............................................................................................................................. 1 Translating SBIRT ................................................................................................................... 3 SBIRT in the United States ............................................................................................................. 4 SBIRT Modifications for Zambia .................................................................................................. 5 Analysis ....................................................................................................................................... 6 Objective .............................................................................................................................................. 6 Standing ............................................................................................................................................... 8 Contribution to the field ................................................................................................................. 9 Parameter Estimates .............................................................................................................. 9 Implementation costs ..................................................................................................................... 9 Patient costs ....................................................................................................................................... 9 Productivity Benefits ..................................................................................................................... 10 Health benefits ................................................................................................................................ 11 Avoided Cost of Crime ................................................................................................................... 12 Social Discount Rate ...................................................................................................................... 13 Figure 1: SBIRT Model ............................................................................................................................... 13 Drinking Categories ....................................................................................................................... 14 Participation and Screening Acceptance Rates .................................................................... 14 Acceptance of In-­βPatient Treatment ........................................................................................ 15 Success Rate ..................................................................................................................................... 15 Model ......................................................................................................................................... 16 Results ................................................................................................................................................ 17 Monte Carlo Results and Analysis ......................................................................................................... 17 Figure 2: Histogram of Net Present Value, Village ......................................................................... 18 Figure 3: Histogram of Net Present Value, Province ..................................................................... 19 Worst-­βCase Scenario ................................................................................................................................... 20 Costs for Lutheran Health Alliance ........................................................................................... 21 Costs and Benefits for Each Screened and Treated Individual ....................................... 21 Costs and Benefits for Each Successfully Treated Individuals ........................................ 22 Average Net Present Value for Population Screened by Sex ............................................ 22 Return on Investment ................................................................................................................................ 23 Conclusion ............................................................................................................................... 23 References ............................................................................................................................... 25 Appendices .............................................................................................................................. 30 Appendix A: Zambian Demographics ....................................................................................... 30 Figure 4: Demographic Characteristics .............................................................................................. 30 Appendix B: Adjusting for the Exchange Rate ....................................................................... 34 Appendix C: Adjusting for Inflation .......................................................................................... 35 iii Appendix D: Calculation of Average Wages in Zambia ...................................................... 36 Appendix E: Calculation of Absenteeism Costs ..................................................................... 38 Appendix F: Calculation of Presenteeism Costs ................................................................... 45 Appendix G: Calculation of Patient Opportunity Cost ........................................................ 52 Appendix H: Calculation of Social network costs ................................................................. 55 Appendix I: Calculation of health care system cost ............................................................. 56 Appendix J: Calculation of avoided costs of crime and road traffic accidents ........... 57 Avoided costs of crime ............................................................................................................................... 57 Avoided costs of road traffic accidents ............................................................................................... 58 Family Cohesion and Domestic Violence ........................................................................................... 60 Appendix K: Implementation Costs .......................................................................................... 61 Salary of Community Health Worker .................................................................................................. 61 Supplies ............................................................................................................................................................ 62 Screening and Brief Intervention .......................................................................................................... 62 Initial and Follow-­βup Training Costs ................................................................................................... 65 Recurrent Training and Support Costs ............................................................................................... 66 Appendix L: Value of a Statistical Life (VSL) .......................................................................... 68 Appendix M: Quality Adjusted Life Years (QALY) ................................................................ 72 Appendix N: Human Immunodeficiency Virus (HIV) .......................................................... 74 Appendix O: Histograms ............................................................................................................... 76 Figure 7: NPV, Province, 2% Discount ................................................................................................ 76 Figure 8: NPV, Village, 2% Discount .................................................................................................... 77 Figure 9: NPV, Province, 6% Discount ................................................................................................ 78 Figure 10: NPV, Village, 6% Discount .................................................................................................. 79 Figure 11: NPV, Province, 10% Success ............................................................................................. 80 Figure 12: NPV, Village, 10% ................................................................................................................... 81 Appendix P: Parameters Summary ........................................................................................... 82 LIST OF TABLES Table 1: Net Present Value, Discount Rate 3.5 percent, Success Rate 20 percent ....................... 17 Table 2: Net Present Value, Discount Rate 3.5 percent, Success Rate 10 percent ....................... 19 Table 3: Net Present Value, Discount Rate 2 percent, Success Rate 20 percent .......................... 20 Table 4: Net Present Value, Discount Rate 6 percent, Success Rate 20 percent .......................... 20 Table 5: Costs Per Program to Lutheran Health Alliance............................................................. 21 Table 6: Costs and Benefits for All Screened ............................................................................... 21 Table 7: Costs and Benefits for Successfully Treated Individuals ............................................... 22 Table 8: Average NPG by Sex ...................................................................................................... 22 Table 9: Demographic Characteristics .......................................................................................... 31 Table 10a: Sex............................................................................................................................... 31 Table 10b: Geographic Location .................................................................................................. 31 Table 10c: Economic Status: Economically Active (includes employed and unemployed) ........ 31 iv Table 10d: Employment Status: Unemployed .............................................................................. 32 Table 10e: Mwembeshi Sector of Employment, Rural ................................................................. 32 Table 11a: Sex............................................................................................................................... 32 Table 11b: Geographic Location .................................................................................................. 32 Table 11c: Economic Status: Economically Active...................................................................... 32 Table 11d: Employment Status: Unemployed .............................................................................. 33 Table 11e: Sector of Employment, Rural ..................................................................................... 33 Table 11f: Sector of Employment, Urban ..................................................................................... 33 Table 12: Zambia Inflation Rates ................................................................................................. 35 Table 13: Female Wage Rates ...................................................................................................... 37 Table 14: Male Wage Rates .......................................................................................................... 37 Table 15: Means days absent attributed to sickness by drinking category in Australia ............... 38 Table 16: Means days absent attributed to sickness by drinking category in Zambia (transferred from South Africa context) ................................................................................................... 40 Table 17: Days absent associated with alcohol consumption per year ......................................... 40 Table 18: Percent of working year absent attributed to alcohol consumption .............................. 41 Table 19: Female Absenteeism Cost Estimates (USD) ................................................................ 43 Table 20: Male Absenteeism Cost Estimates (USD) .................................................................... 44 Table 21: Sick-related Absenteeism Rate ..................................................................................... 46 Table 22: Adjusting US Presenteeism Days to Zambia the SBIRT Likely Dependent drinking . 46 Table 23: Ratio of days absent between drinking states ............................................................... 47 Table 24: Days presenteeism occurs associated with alcohol consumption per year ................... 47 Table 25: Percent of working year presenteeism occurs attributed to alcohol consumption........ 48 Table 26: Female Presenteeism Cost Estimates (USD) ................................................................ 49 Table 27: Male Presenteeism Cost Estimates (USD) ................................................................... 50 Table 28: Cost of Crime................................................................................................................ 57 Table 29: Cost of road traffic accidents ........................................................................................ 59 Table 30: SBIRT Costs per Drinker.............................................................................................. 64 Table 31: Training Costs............................................................................................................... 66 Table 32: Values Used in VSL Calculation .................................................................................. 69 Table 33: Values Used in VSLY Calculation ............................................................................... 70 Table 34: Estimates of change in QALY after SBIRT treatment ................................................. 73 Table 35: HIV Acquisition Odds by Drinking State ..................................................................... 74 Table 36: Parameters Treated as Uncertain .................................................................................. 82 Table 37: Parameters Treated as Certain ...................................................................................... 85 v EXECUTIVE SUMMARY This cost-benefit analysis was prepared at the request of the Lutheran Health Alliance (LHA) to examine the value of implementing a Screen, Brief Intervention and Referral to Treatment (SBIRT) program in Zambia. The purpose of SBIRT is to prevent and reduce harmful alcohol consumption patterns. The SBIRT program analysis first details the implementation of the pilot program in Mwembeshi, a rural community in Central Province, Zambia. We also consider the costs and benefits associated with scaling the program up to the level of the Central Province. Based on our modeling we recommend the implementation of the pilot program. However, because the results are sensitive to the success rate of the program, we also recommend that LHA monitor the success rate of the pilot program before scaling up to the provincial level. We incorporate various costs and benefits that accrue to LHA, the individual, or society. Cost categories include the following: implementation costs; social network costs; and patient opportunity costs. Benefit categories encompass health benefits; social benefits; productivity benefits; and reduction in health care system benefits. Translating these costs and benefits into a Zambian context required researching Zambian statistics, and when no appropriate statistics were available, making reasonable assumptions for predicting the impacts of SBIRT in Zambia. This framework yielded positive net benefits for both the village and province level programs. In Mwembeshi, the net present value for a successfully treated individual is $2,530 For the Central Province, these figures is $2640. Taking a broader perspective, the estimated net present value of the program overall for Mwembeshi was $202,200 and $35,947,000 for Central Province. Net benefits remained positive using varying discount rates and a lower success rate. vi Based on these findings we recommend that the pilot program be implemented. The evidence suggests SBIRT is an effective intervention for changing alcohol consumption patterns. Moreover, with the pilot program using, in large part, existing resources, little costs are incurred in comparison to the benefits gained. In this respect our results show that for every dollar of investment, there is at least a ten-dollar return in benefits. Before scaling up to the provincial level, however, we suggest monitoring the results of the pilot program. In particular, we suggest reviewing the participation rates, the success rate, and the length of time the change in alcohol consumption is sustained. By doing a field experiment, the accuracy of the model for the province can be significantly improved, and thus, the appropriate level of investment will be better informed vii INTRODUCTION In Zambia, high alcohol consumption, both in terms of quantity and frequency imposes harm on people. Several sources estimate the prevalence of alcohol consumption in Zambia. The World Health Organization’s Global Status on Health and Alcohol Report finds that 2.2 percent of the total Zambian population engages in heavy episodic drinking (WHO 2014). The prevalence is heavily weighted toward men, 4.4 percent of men and less than .01 percent of women are classified as heavy episodic drinkers. However, the WHO estimates do not include unrecorded consumption, resulting in an underestimated prevalence of heavy drinking. Based on Dr. Jason Paltzer’s qualitative research, alcohol substance abuse is prevalent in 20 to 50 percent of the population. To put this statistic in a more local perspective, the estimated drinking prevalence in Zambia is twice that in Wisconsin. Other studies state that around 15 percent of the total population can be classified as heavy drinkers (Haworth 2004). We estimate a distribution of the various drinking states by gender to account for different drinking patterns among men and women. Cultural attitudes support a heavy drinking culture in Zambia. Heavy drinking episodes are common practice with cultural and traditional ties accompanied by a high degree of social acceptance in Zambian society. On one end of the spectrum there are those who abstain from alcohol consumption, on the other end, are those who consume alcohol in excess. Consistent with Dr. Paltzer’s research, Allan Haworth reports, “The close correlation between responses about drinking and those about drunkenness suggests to us that the norm among Zambian drinkers is that one goes drinking in order to get drunk” (Haworth 2004). Moreover, there is a negative perception of those who abstain from drinking or those who moderately drink. For 1 example, abstainers or moderate drinkers may be labeled a “womanizer,” a “cheat,” or “a thief” (Paltzer 2009). Being regarded as such is considered for many Zambians worse than being labeled a drunkard (Paltzer 2009). Additionally, the drinking culture appears to be largely uniform, with little difference based on educational attainment, type of job, or occupational status relative to the community as a whole (Haworth 2004). Zambia faces challenges in the capacity of its health care infrastructure for combating heavy episodic drinking, especially when confronted with a high HIV prevalence rate. The country has a high burden of disease especially in terms of communicable diseases such as malaria, tuberculosis (TB), human immunodeficiency virus (HIV), and sexually transmitted infections (STIs). The HIV prevalence rate is 12.5 percent, which is considered at a national epidemic level (AVERT 2012). Further compromising the effectiveness of the health care system are limited financial resources, “weak logistics management in the supply of drugs and medical supplies” (Zambia’s Ministry of Health 2011), outdated medical equipment, and a “brain drain” of medical professionals. The government has only about half of the health care workers it seeks. For example, according to a report by Joseph Schatz, Zambia has fewer than 646 doctors and 6096 nurses, representing only 28 percent and 36 percent of the government’s target (Schatz 2008). To put this in relative terms, in Zambia there is one doctor and seven nursing personnel per 10,000 people compared to 27 doctors and 98 nurses per 10,000 people in the United States (Global Health Workforce Alliance 2014). Non-governmental organizations are working to improve the health status of Zambians by piloting health interventions at the community level. One such organization is the Lutheran Health Alliance (LHA), a subsidiary of the Kingdom Workers, which serves churches and communities by providing support for volunteer programs internationally. LHA has sent 2 professional medical trainers to work with communities in Zambia to attend to public health needs. In addition, LHA maintains a partnership with the Lutheran Health and Development Program located in Zambia, which prepares volunteers for service by training them in health education, disease prevention, and responding to basic health care needs. The Lutheran Health and Development Program consists of a manager and a program assistant who are responsible for the training of volunteers and the coordination of the program in communities. Specific to Mwembeshi, Zambia, LHA works with 25 lay community health workers (CHWs) to provide health education, treat malaria and other common ailments, and refer patients to professional medical care as needed. CHWs are becoming increasingly vital in addressing the scarcity of healthcare services while bridging the divide between the professional workforce and the surrounding communities. The WHO defines CHWs as “community health aides selected, trained and working in the communities from which they come” (Ashraf and Kindred 2011). CHWs are utilized in communities around the world to meet health needs; however, the role of CHWs and the training they receive varies widely. Generally, CHWs in less formalized programs work in health promotion and educational activities and may receive non-monetary and in-kind donations from the community. Programs formalized through government agencies often provide wages and more extensive training to full-time CHWs (Ashraf and Kindred 2011). TRANSLATING SBIRT One intervention for addressing the costs stemming from alcohol abuse is Screening, Brief Intervention, and Referral to Treatment (SBIRT). SBIRT programs seek to identify people with harmful or hazardous alcohol consumption before the health and social consequences of their 3 drinking become pronounced and motivate them to address their real or potential problem (WHO 2014). In the subsequent sections, we discuss SBIRT in the context of the United States and modifications needed for it to be effective in the Zambian context. SBIRT IN THE UNITED STATES In the United States, Wisconsin has the highest rate of binge drinking and persistent rates of alcohol abuse and misuse (WDHS 2012). Economic and health consequences to the state are numerous, yet state residents have one of the lowest risk perceptions for binge drinking in the country (WIPHL 2012). To address this challenge, the Substance Abuse and Mental Health Services Administration (SAMHSA) awarded the Wisconsin Department of Health Services a five-year grant to implement a SBIRT program in the state. This initiative, entitled Wisconsin Initiative to Promote Healthy Lifestyles (WIPHL), sought to improve the health and lives of residents by including SBIRT in the routine care given by primary providers. The SBIRT protocol begins with a brief screen, typically administered as a written questionnaire at patient check-in. The screen consists of four items to rapidly determine patient experiences with alcohol or drugs within the last three months. Medical assistants score the screen and those considered “positive” are guided through a full screen to determine their level of alcohol use and risk (low risk use, risky use, harmful use, or likely dependent use). The four potential risk levels elicit to distinct interventions. Patients determined to have risky use are led through a brief intervention using motivational interviewing techniques. Those thought to have harmful use are given a brief treatment and those who show likely dependent use are referred to more extensive treatment. Clinical staff often have no more than 10 to 15 minutes with a patient to complete the full screen and planned service (WIPHL 2012). 4 The WIPHL initiative showed a 20 percent reduction of risky alcohol use among patients who participated in the outcome study, which when viewed from a population level, translates to substantial reductions in emergency department visits, hospitalizations, arrests, and vehicular crashes. The results in Wisconsin were similar to those in other states, which also found SBIRT can be both effective and efficient, and that it likely saves money and lives (WIPHL 2012). SBIRT MODIFICATIONS FOR ZAMBIA The primary points of departure between the Zambian and Wisconsin models are the health care setting, the level of expertise of the workforce, and the representativeness of the population reached. In the WIPHL model, the screening occurs during check-in at hospitals and clinics for those seeking care. The subsequent steps are implemented by doctors and nurses at the same, or in a similar, health care facility. This model is not feasible to replicate in Zambia because of the critical shortage of health care professionals. To overcome this low density CHWs provide a significant amount of health care services in Zambia. Adapting SBIRT from a clinical to community setting would shift the screening and intervention stages to the patients’ homes. CHWs can be trained to offer the screening and brief intervention components, and then make referrals in cases of dependency to nearby health centers. This reduces the burden on clinics and broadens the scope by allowing the substance abuse intervention to reach more people. In the case of Mwembeshi, individuals would be referred to the Serenity Harm Reduction Programme Zambia clinic in Lusaka, Zambia (SHARPZ). In a feasibility study of a cognitive behavioral treatment to reduce alcohol consumption among HIV-infected outpatients of a clinic in Kenya, content changes improved the cultural applicability of the program (Papas et al. 2010). Similar to qualitative reports on Zambia, this study reported little prior knowledge regarding the risks and harmful effects associated with 5 alcohol, misinformation about traditional brews, and the importance of drinking in social circles. In response to these observations, a risk education module was added to the first counseling session, as well as an emphasis on alcohol refusal skills and a discussion of the financial costs of drinking in the interventions. We assume the tailoring of the program to these motivations for change would be able to increase the effectiveness of SBIRT for Zambians. Several studies point to CHWs being effective at increasing access, promoting knowledge of health maintenance and disease preventive, and encouraging behavioral change (Swider 2002). Such evidence influenced our assumptions regarding the participation and success rates, as well as our calculations of the implementation and patient costs. Monetization of costs and benefits was another significant difference. All measures were calculated using Zambian wage, inflation, and cost data. Other potential differences were considered and are discussed in terms of each cost and benefit category affected. ANALYSIS OBJECTIVE The primary purpose of the following cost-benefit analysis is to determine the value of expanding an existing CHW program to include SBIRT. This expansion includes training an existing cohort of 25 CHWs to have the skills necessary to implement SBIRT in the community where they serve. This augments the set of health interventions and services they currently provide, instead of displacing these activities. All potential costs and benefits resulting from this change were then considered and evaluated for a rural community in Central Province. Our secondary objective is to determine the value of scaling up the current program, with the inclusion of SBIRT services, to all of Central Province. In making this determination, we 6 considered differences in wages for the CHWs, differences in administrative costs, and differences in the population served. With the shortage of health care professionals and observations of unsafe alcohol consumption patterns, there exists opportunities for large economic and health benefits resulting from successfully implementing substance abuse interventions throughout Zambia. Despite this potential, research in this area stresses that addressing the administrative and capacity requirements for scaling up a health intervention should be done alongside consensus building initiatives (Nyonator et al. 2005). Thus, dialogue with community leaders and the inclusion of residents in the decision-making process is vital when seeking to scale up such a program. LHA is interested in including a mobile-based platform to SBIRT in Zambia. A mobile phone platform could reduce the face-to-face interaction with a health worker and could receive information via short message services. The rapid growth of mobile phone use in low-income countries has opened the possibility to implementing this technology in health care. In Zambia, between 2004 and 2009, the number of mobile phone subscribers rose from just 464,000 to some 4.4 million (Montez 2010). The use of mobile phones in health-related fields has been applied in other countries already and research shows positive results from their use such as improving health workers’ adherence to malaria guidelines in Kenya (Zurovac et al. 2012) and delivering mental health interventions in South Africa (Norris et al. 2013). Notwithstanding, the country faces several drawbacks to implementing such technology. Zambia does not consistently meet electricity demand and many parts of the country lack grid electricity. Consequently, those who own a cell phone often must travel to other towns to pay to charge their cell phones. In addition, phone lending is a common practice in Zambia, especially in rural areas. Results from a survey show that “26 percent of mobile phone owners said they 7 lend their phone to someone else at least once a month” (Montez 2010). Besides the resource constraint, the implementation of this type of intervention is fairly new in the country and the region, which makes it difficult to determine how feasible and effective such technology would be to encourage participation and improve adherence to SBIRT from patients. We anticipate that adding a mobile phone component would add a fixed cost for each community health worker and add to the training costs, with the potential to increase the effectiveness rate. However, due to uncertainty of the change in effectiveness to substance abuse intervention in the specific communities, we concluded that the results would be more informative without this added complexity, and omitted analysis of a mobile-based alternative. STANDING We assign standing to LHA, and all individuals 15 or older within the geographic area of interest. LHA would incur the implementation costs. These costs would change as the program is expanded from the village to provincial level. Society would incur both costs and benefits. All individuals in society may incur benefits and costs within the treated geographic area. At the village level and province level we assign costs and benefits to every individual aged 15 and older. Costs and benefits vary across drinking states, gender, income status, and geographic area. Although evidence points to the occurrence of drinking among adolescents in Zambia, our analysis did not incorporate this age group because a significant portion of our data was based on the population 15 years and older. Additionally, for those 15 years and younger, benefits would have to include the monetization of uncertain educational and mental health benefits. Therefore, we believe we conservatively estimate overall benefits. Additionally, those under the age of 15 are outside the scope of the SBIRT implementation and will not incur any of the direct costs associated with the screening, brief intervention, and referral to treatment. 8 CONTRIBUTION TO THE FIELD Our cost-benefit analysis presents a new perspective on estimating the effects of the SBIRT model in a developing context. To date almost all cost-benefit analyses of SBIRT remain in high-income industrialized countries. Both Maltzopoulos et al. (2014) and Thavorncharoensap et al. (2010) attempt to monetize the costs of excessive alcohol consumption in South Africa and Thailand, respectively. However, the developing context lacks a comparative comprehensive cost-benefit analysis that examines the costs and benefits associated with reductions in excessive problem drinking. We believe our report will assist in bridging the information gap and provide a foundation for future cost-benefit analysis in the low-income and sub-Saharan African context. PARAMETER ESTIMATES IMPLEMENTATION COSTS The costs of implementing SBIRT include the initial training of CHWs, their wages for implementing the screenings and interventions, as well as the supplies needed for these activities. CHWs receive an initial training of approximately two weeks immediately and receive a multiday, follow-up training after two years. In the pilot phase, employees of the Lutheran Health and Development Program serve as the primary trainers. An additional implementation cost is the clinical support costs when CHWs consult with local clinical staff. More elaboration on the calculation of these implementation parameters can be found in Appendix K. PATIENT COSTS This category includes both costs and avoided costs (benefits) directly resulting from the patient having completed the alcohol abuse screening and intervention process. The most direct of these cost measures is the opportunity cost for the time an individual spends receiving a brief 9 screen, brief intervention, or treatment, as applicable for the designated risk classification. Further explanation of this calculation can be found in Appendix G. Another cost borne by the individual receiving the health service is social network costs. This is included in the analysis given the high importance referrals have in attaining employment and the connectedness of drinking and peer groups. Further explanation of this cost can be found in Appendix H. The final measure of this cost category examines changes in health care utilization resulting from SBIRT. Based on research in this area, the resulting change is expected to be two-fold. First, the education that goes alongside SBIRT increases awareness of health care needs, which in turn increases the utilization of outpatient services (Paltzer 2014). At the same time, anticipated decreases in alcohol-attributable diseases would decrease the amount of inpatient and ER services used (Paltzer 2014). The exact calculations for this category of patient costs can be found in Appendix I. PRODUCTIVITY BENEFITS Productivity benefits include the benefits derived from increased worker productivity resulting from the reduction in alcohol consumption and frequency. We articulate two types of productivity benefits that are realized by the individual drinker, the employer, and the greater community. First, we calculate absenteeism benefits, which are associated with reduced workdays attributed to alcohol missed for the employed individual. We follow the methodology used by Quanbeck et al. (2010) that calculates absenteeism by multiplying the daily wage by the days absent associated with alcohol consumption (Quanbeck et al. 2010; Baumberg 2010). We translate days absent into a percentage of one’s working year. Our final calculation is the percentage of the working year absent attributed to alcohol multiplied by the annual wage and a 10 spillover factor. Dollar value estimates vary based on drinking categories and wages. Wages are conditional on geography, sector of employment, and gender. Appendix E shows the calculation of absenteeism costs under current policy. Second, we calculate presenteeism benefits, which are associated with reduced impaired productivity at work. Impaired productivity may take the form of arriving drunk to work or being hung over at work (Maltozpoulos et al. 2014). We use data the Arora et al. (2011) estimates of the range of days that a problem drinker experiences presenteeism. Using the likely dependent drinker as the base case, we estimate presenteeism for each drinking state by applying the same rate at which absenteeism changes between drinking states. Similar to absenteeism we translate the number of days presenteeism occurs into a percentage of one’s working year. Our final calculation is the percentage of the working year presenteeism occurs that is attributed to alcohol multiplied by the annual wage and an impairment factor (Quanbeck et al. 2010; Thavorncharoensap et al. 2010). Estimates vary based on drinking categories and wages. Wages are conditional on geography, sector of employment, and gender. Appendix F shows the calculation of presenteeism costs under current policy. HEALTH BENEFITS Heavy alcohol consumption is associated with increased risk of several diseases and negative health conditions. Drinking directly causes some conditions, such as fetal alcohol syndrome and alcohol dependence. Others, such as slips, falls, and accidental injuries are caused by alcoholic impairment (Rehm et al. 2003). Other conditions are partially caused by, or correlated with alcohol consumption, such as esophageal cancer, liver cirrhosis, heart disease, hypertension, and diabetes. Reducing drinking therefore also reduces the risk of acquiring one or more of these conditions. For detailed calculations see Appendices L,M, and N. 11 Increases in the quality of life are measured in quality-adjusted life years, where a year of perfect health is rated as 1.0 and reduced proportionate to the reduction in quality of life. Successful treatment of SBIRT is found to enhance a patient's life by .091 to .405 QALYs per year, depending on sex, drinking level, and course of treatment. AVOIDED COST OF CRIME We measure the costs of crime that are attributed for each drinker as the avoided costs per individual who is successfully treated. This measure includes the cost to all of society, but is only attributed to those who drink. This is because alcohol-attributable crime will only be reduced for those who presently consume alcohol. We also consider that by reducing alcohol consumption an individual will reduce the probability of dying in an alcohol-related crime incident. Based on studies of the costs of crime in the United States (Boucher et al. 2006) and South Africa (Matzopoulos et al. 2014) and adjusting the estimates for Zambia, a lower income level country with lower crime rates than in these studies, we estimate that 50 percent of costs of crime are alcohol-related using an alcohol-attributable fraction (AAF) of 0.25. However, given that not all crimes are reported and we lack official data from Zambia, our numbers for crime are underestimated. AVOIDED COSTS OF ROAD TRAFFIC ACCIDENTS Another benefit category associated with the implementation of SBIRT includes the avoided costs for victims of traffic accidents. These costs as well as the reduction in the probability of dying in a road traffic accident are attributable to drinkers. Using estimates from the United States, South Africa, and Thailand as points of reference, alcohol-related traffic accidents in Zambia cost between 0.25 and 1 percent of GDP considering that Zambia is a lowermiddle income country and has lower number of motor vehicles registered than the reference 12 countries (see Appendix J). These costs include damages in crash, lost of productivity, lost of lives, and correction costs associated with alcohol. SOCIAL DISCOUNT RATE We discount all costs and benefits according to the following specified social discount rates. First, we assume a real discount rate of 3.5 percent. Second, we check the robustness of our results with lower bound real discount rate of 2 percent and an upper bound real discount rate of 6 percent. We use these three discount rates according to the justifications and recommendations set forth by Moore et al. (2004). FIGURE 1: SBIRT MODEL Source: Authors 13 Figure 1 maps the SBIRT flowchart and provides assumptions about probabilities of participation. DRINKING CATEGORIES The drinking categories as designated by SBIRT include low-risk, at-risk, harmful, and likely dependent. To estimate the distribution of drinkers for both the pilot and province level projects, we use a 2008 study conducted by the Zambian Ministry of Health in Lusaka prevalence rates of risk factors for noncommunicable diseases. Survey responses provided estimates for males and females who abstain from drinking (56.5 percent and 82.4 percent, respectively), and the frequency and quantity of drinking for those that consume alcohol. Among drinkers, 9 of 10 men responded that they consume more than three to four drinks on average per day, and 7 out of 10 women reported drinking more than three to four drinks per day. This consumption pattern alone would most likely qualify them for the at-risk or harmful drinking categories. Our drinking categories are based on the frequency of drinking considering that the drinking culture is dichotomous; with a high rate of abstainers but high rates of excessive drinking among those who do consume alcohol. The percentage of drinkers in each category in our model are as follows: for men, 12.5 percent low-risk, 31.4 percent at risk, 39.9 percent harmful, and 19.2 percent likely dependent; for women, 34.7 percent low-risk, 36.9 percent at risk, 20.9 percent harmful, and 7.5 percent likely dependent. PARTICIPATION AND SCREENING ACCEPTANCE RATES We expect that not all adults aged 15 and older will participate with the CHWs, and of those who participate not all will accept the initial screen. Dr. Paltzer indicated that anywhere from 50 to 80 percent of families currently participate with CHWs in Mwembeshi. We assume this participation rate would be similar for the SBIRT implementation and we take the midpoint 14 of 65 percent participation into account when estimating the target population. In addition, from a study on home-based voluntary HIV counseling and testing in Zambia, 85 percent of HIV positive individuals working with CHWs accepted the intervention (Fylkesnes et al. 2013). We use this number to estimate the number of adults who will work with CHWs and accept the initial screen. ACCEPTANCE OF IN-PATIENT TREATMENT We included an additional probability to account for the anticipated lower participation of the likely dependent due to greater opportunity costs. All of those who are classified as likely dependent receive the brief and full screens, and so, all in this category have the opportunity costs associated with those two stages. However, those who accept the referral and go to a nearby clinic to receive treatment face the further costs associated with traveling, waiting, and the nontrivial amount of time spent in treatment. The acceptance rate we used to differentiate those who accept treatment and those who go no further than the full screen comes from a systematic review of patient retention in antiretroviral therapy programs throughout sub-Saharan Africa. This study finds a 70 percent follow-up retention rate after 24 months of first receiving treatment (Fox and Rosen 2010). While not directly tied to Zambia or alcohol interventions, this provides the closest statistic available, which enables us to factor in this crucial probability. To improve the likelihood of having a high acceptance rate, LHA could reimburse individuals in this drinking category for the travel costs, to reduce the individual burden for treatment. SUCCESS RATE We incorporate a success rate of 20 percent, meaning of all the individuals who are screened, receive brief interventions, or are referred to treatment; one in five will successfully change their drinking behavior. This is supported in evaluations of SBIRT in Wisconsin 15 (SCAODA 2013) among others. Because of the differences in the drinking culture between Zambia and the United States, we do not know the actual success rate that will be seen in the pilot program, and so we use this relatively conservative estimate for our analysis. MODEL Many of the costs and benefits that we have included in our analysis are uncertain. For example, it is not known exactly how many individuals abstain from drinking. To account for uncertainty, we chose to create and implement a Monte Carlo simulation, a computer model that integrates uncertain or unknown values over a large number of independent simulations. First, we have selected point estimates for parameters needed to calculate costs and benefits from existing research and data sources. For all uncertain parameters, we also specify a distribution or range of values. For the majority of our parameters, we have chosen a triangle distribution centered on our point estimates, which become the most likely value used in calculations. Second, we use a computer model to select a random value from each distribution, and then calculate the costs and benefits that accrue to an individual receiving screening and treatment, LHA, and society at large. We then repeat this process 10,000 times to obtain an average value of net benefits. Initially, our model simulates a random sample of 10,000 individuals. Certain characteristics, including gender; drinking category; labor sector; urban-rural classification; and treatment participation, acceptance, and success are independently calculated based on demographic estimates of the Zambian population. The model then assigns values to remaining parameters based on these characteristics and the specified distribution. See Appendix Q for parameter values and range distributions. 16 The majority of costs and benefits are calculated per individual, then multiplied by the number of participants in both the village and province. We calculate the costs and benefits of this project over a four-year time horizon. At year zero, initial CHW training costs and the costs of implementing the screenings and brief interventions are incurred. Social network costs are calculated at the midpoint of year one, and follow-up CHW training costs are incurred at the midpoint of year two. The remaining costs and benefits are dependent upon successful SBIRT treatment. Past SBIRT evaluations have shown varying duration of program effects. We have taken a conservative approach and have built a linear decline of effects over the time horizon of the program and assume effects return to baseline at the end of year four. RESULTS MONTE CARLO RESULTS AND ANALYSIS After executing the Monte Carlo analysis and using population estimates to calculate the net present values (NPVs) of the costs and benefits of the overall program, we find that the average NPV for the pilot is $200,200 and the average NPV for the program scaled to the province level is $42,297,000. All simulations resulted in positive NPVs on both the pilot and province levels. Table 1: Net Present Value, Discount Rate 3.5 percent, Success Rate 20 percent Program Average NPV Lower Bound Upper Bound Pilot $202,200 $171,000 $231,500 Province $35,947,000 $30,869,000 $42,297,000 Source: Authors 17 FIGURE 2: HISTOGRAM OF NET PRESENT VALUE, VILLAGE Source: Authors 18 FIGURE 3: HISTOGRAM OF NET PRESENT VALUE, PROVINCE Source: Authors Next, we calculated the NPV for both the pilot and province level programs assuming a success rate of SBIRT of 10 percent. While the average NPV of the benefits was reduced, all simulations resulted in positive net benefits. Table 2: Net Present Value, Discount Rate 3.5 percent, Success Rate 10 percent Program Average NPV Lower Bound Upper Bound Pilot $115,500 $93,400 $141,600 Province $21,112,000 $17,357,000 $25,033,000 Source: Authors 19 Additionally, we varied the discount rate from 3.5 percent to a lower bound of 2 percent and an upper bound of 6 percent. Overall estimates were sensitive to discount rate, primarily due to increased value of a statistical life year (see Appendix L.) Table 3: Net Present Value, Discount Rate 2 percent, Success Rate 20 percent Program Average NPV Lower Bound Upper Bound Pilot $163,300 $137,900 $186,600 Province $29,076,000 $285,244,000 $33,897,000 Source: Authors Table 4: Net Present Value, Discount Rate 6 percent, Success Rate 20 percent Program Average NPV Lower Bound Upper Bound Pilot $273,600 $232,000 $314,500 Province $48,621,000 $41,248,000 $57,795,000 Source: Authors WORST-CASE SCENARIO To further verify the sensitivity of our results to assumptions made in our model we performed an analysis of a worse case scenario. This worse case scenario used statistics far below any suggested by previous research. This modeling used a six percent discount rate, a five percent success rate, and a 50 percent participation rate. Even with these very low parameters, the outcomes of the programs are still positive in 100 percent of the trials of the Monte Carlo simulation. The resulting average net present value for Mwembeshi is $51,100 and is $12,474,000 for the Central Province. 20 COSTS FOR LUTHERAN HEALTH ALLIANCE Costs to implement SBIRT for LHA are modest, particularly for the pilot in Mwembeshi. The changes between the pilot and province levels result primarily from the formalization of the CHW workforce, and administrative overhead costs. Table 5: Costs Per Program to Lutheran Health Alliance Cost/Benefit Category Mwembeshi Central Province Training Costs $6,000 $1,039,000 Clinic Support Costs $11,500 $185,000 Program Costs $2,800 $733,000 Overhead Costs 0 $450,000 Total Costs $20,300 $2,407,000 Source: Authors COSTS AND BENEFITS FOR EACH SCREENED AND TREATED INDIVIDUAL Patient opportunity costs increase from the pilot to province level based on the change in the make-up of the workforce, with a higher percentage of individuals living in urban areas and working in the formal sector in Central province Table 6: Costs and Benefits for All Screened Cost/Benefit Category Mwembeshi Central Province Patient Opportunity Costs $1.10 $1.50 Health System Benefits $39.70 $39.70 Source: Authors 21 COSTS AND BENEFITS FOR EACH SUCCESSFULLY TREATED INDIVIDUALS For successfully treated individuals, the only cost apart from the opportunity and travel costs is the loss of their social network, a one-time cost calculated at the midpoint of year one. The net benefits at the pilot level are $1005 and $907 at the province level. The primary driver of the positive benefits is accumulated health benefits and in particular the benefits associated with reduction in risk of HIV infection with reduced drinking. Table 7: Costs and Benefits for Successfully Treated Individuals Cost/Benefit Category Mwembeshi Central Province Social Network Costs $90 $190 Health Benefits $700 $700 HIV Reduction Benefits $1,610 $1610 Absenteeism $20 $25 Presenteeism $1 $5 Crime $65 $65 Traffic Accidents $45 $45 Overall Net Present Value $2,530 $2640 Source: Authors AVERAGE NET PRESENT VALUE FOR POPULATION SCREENED BY SEX We found the average net benefits for both men and women to be higher in Mwembeshi than in Central Province. This difference can be attributed to the lower implementation costs at the village level. Table 8: Average NPG by Sex Individual Average Net Present Value 22 Men (Mwembeshi) $237 Women (Mwembeshi) $112 Men (Central Province) $231 Women (Central Province) $107 Source: Authors RETURN ON INVESTMENT When SBIRT has been systematically implemented with fidelity to the model, it has been shown to produce a return on investment of four to one within the first year of implementation (SCAODA 2013). The primary costs of implementing SBIRT in the United States are the wages and benefits of the health worker or clinician giving the screenings and interventions. Modifying the model to use CHWs dramatically reduces implementation costs. Correspondingly, our model estimates a return on investment of approximately 10 to one over the course of four years. We did total benefits by total costs to calculate return on investment. CONCLUSION We recommend LHA implement the SBIRT pilot program in Mwembeshi, Zambia. Finding positive net benefits occurring in 100 percent of our trials, considering both a 20 percent and a 10 percent success rate, we believe implementing SBIRT would be an efficient use of resources. With an initial investment in alcohol screening and intervention, numerous benefits result, extending into other aspects of society, including benefits in income, safety, health, societal cohesion. Despite overwhelming positive results, there are several notable limitations and assumptions that call for prudence when interpreting and implementing our findings. 23 One of these limitations is that our model relies heavily on data taken from a handful of studies and various assumptions stated throughout our analysis. Thus, the validity of our model depends greatly on the reliability and validity of these studies. In many cases this means data from the Zambian government and the WHO. For other statistics, including the acceptance of screening rate and acceptance of referral to treatment, there is additional uncertainty because our model uses data for HIV treatment and assumes alcohol abuse treatment will be comparable. To gauge the appropriate level of the investment we also recommend that the LHA monitor the pilot program before scaling up to all of Central Province. The response of the Zambian communities to the SBIRT program is entirely unknown; particularly the success rate and the time horizon of benefits. This program is seeking to change Zambia’s drinking culture by encouraging moderate alcohol consumption by reducing harmful alcohol use. Any intervention seeking to instigate such a behavior change with an addictive substance is highly uncertain. As such, a reevaluation will be needed to assess the appropriateness of our assumptions and probabilities used. With that said, all of our results point to SBIRT being a high-yield investment, leading to substantial benefits to individuals and society overall. 24 REFERENCES Agerwala, Suneel M., and Elinore F. McCance-Katz. 2012. “Integrating Screening, Brief Intervention, and Referral to Treatment (SBIRT) into Clinical Practice Settings: A Brief Review.” Journal of Psychoactive Drugs 44 (4): 307–17. Alexander Forbes Health Management Solutions. 2013. Alexander Forbes Health Management Solutions Calls for pro-Active Sick Management Strategy in South Africa. Anderson, Peter. 2010. Alcohol and the Workplace: A Report on the Impact of Work Place Policies and Programmes to Reduce the Harm Done by Alcohol to the Economy. Arora, Prachi, Elizabeth Hartjes, Nathaniel Inglis-Steinfeld, Noah Natzke, and Alissa Quade. 2011. Screening, Brief Intervention, and Referral to Treatment Cost-Benefit Analysis for Wisconsin Medicaid Patients. Prepared for Jason Paltzer, Population Health Institute School of Medicine and Public Health University of Wisconsin – Madison. December 22 Ashraf, Nava, and Natalie Kindred. 2011. “Community Health Workers in Zambia β―: Incentive.” AVERTing HIV AND AIDS (AVERT). 2012. “HIV and AIDS Statistics 2011.” http://www.avert.org/africa-hiv-aids-statistics.htm. Baumberg, Ben. 2010. Best Practice in Estimating the Costs of Alcohol – Recommendations for Future Studies. Boardman, Anthony E., David H. Greenberg, Aidan R. Vining, and David L. Weimer. Cost-benefit Analysis. Routledge, 2007. Bouchery, Ellen E, Henrick J Harwood, Jeffrey J Sacks, Carol J Simon, and Robert D Brewer. 2011. “Economic Costs of Excessive Alcohol Consumption in the U.S., 2006.” American Journal of Preventative Medicine 41 (5). Elsevier Inc.: 516–24. Bray, Jeremy W., Gary A. Zarkin, Jesse M. Hinde, and Michael J. Mills. 2012. "Costs of Alcohol Screening and Brief Intervention in Medical Settings: A Review of the Literature." Journal of Studies on Alcohol and Drugs: 911-19. Bureau of Labor Statistics (BLS). 2011. Absenteeism Household Data. Central Statistical Office (CSO), Ministry Labour and Social Security, Zambia. 2008. Labour Force Survey Report 2008. ———. 2013. Preliminary Results of the 2012 Labour Force Survey. Central Statistical Office (CSO), Ministry of Health (MOH), Tropical Diseases Research Centre (TDRC), and and Macro International Inc University of Zambia. 2009. Zambia Demographic and Health Survey 2007. Calverton, Maryland. Chola, Lumbwe, and Bjarne Robberstad. 2009. “Estimating Average Inpatient and Outpatient Costs and Childhood Pneumonia and Diarrhea Treatment Costs in an Urban Health Centre in Zambia.” Cost Effectiveness and Resource Allocation β―: C/E 7 (January): 16. Duvvury, Nata, Caren Grown, and Jennifer Redner. 2004. Costs of Intimate Partner Violence at the Household and Community Levels for Developing Countries. Farnham, Paul G, David R Holtgrave, Chaitra Gopalappa, Angela B Hutchinson, and Stephanie L Sansom. 2013. “Lifetime Costs and Quality-Adjusted Life Years Saved from HIV 25 Prevention in the Test and Treat Era.” Journal of Acquired Immune Deficiency Syndromes (1999) 64 (2): e15–18. Fleming, Michael F, Marlon P Mundt, Michael T French, Linda Baier Manwell, Ellyn A Stauffacher, Kristen Lawton Barry, and Care. 2000. “Benefit-Cost Analysis of Brief Physician Advice with Problem Drinkers in Primary Care Settings.” Medical Care 38 (1): 7–18. Fox, Matthew P, and Sydney Rosen. 2010. “Patient Retention in Antiretroviral Therapy Programs up to Three Years on Treatment in Sub-Saharan Africa, 2007-2009: Systematic Review.” Tropical Medicine & International Health β―: TM & IH 15 Suppl 1 (june): 1–15. Fretz, Caley. 2011. “World Bicycle Relief Utility Bikes: Not UCI-Legal, but the Better for It.” http://velonews.competitor.com/2011/01/bikes-and-tech/world-bicycle-relief-utility-bikesnot-uci-legal-but-the-better-for-it_156354#i6WwXvleQLVUYcvv.99. Fylkesnes, Knut, Ingvild Fossgard Sandøy, Marte Jürgensen, Peter J Chipimo, Sheila Mwangala, Charles Michelo, and Ingvild Fossgard. 2013. “Strong Effects of Home-Based Voluntary HIV Counselling and Testing on Acceptance and Equity: A Cluster Randomised Trial in Zambia.” Social Science & Medicine (1982) 86 (June): 9–16. Gentilello, Larry M., Beth E. Ebel, Thomas M. Wickizer, David S. Salkever, and Frederick P. Rivara. 2005. “Alcohol Interventions for Trauma Patients Treated in Emergency Departments and Hospitals.” Annals of Surgery 241 (4): 541–50. Global Health Workforce Alliance. 2011. “Which countries are faced with a ‘critical’ health workers shortage?” The Guardian, 17 January 2011. October 14, 2014. http://www.theguardian.com/global-health-workers/interactive/infographic-mortatlity-rateshealth-workers-uk-us-worlld Hämäläinen, Päivi, Jukka Takala, and Kaija Leena Saarela. 2006. “Global Estimates of Occupational Accidents.” Safety Science 44 (2): 137–56. Haworth, Allan. 2004. “Local Alcohol Issues in Zambia.” In Moonshine Markets, 41–66. Hjortsberg, Catharina. 2003. “Cost of Access to Health Services in Zambia.” Health Policy and Planning 17 (1): 71–77. ———. 2003. “Why Do the Sick Not Utilise Health Care? The Case of Zambia.” Health Economics 12 (9): 755–70. Hutchinson, Angela B, Pragna Patel, Stephanie L Sansom, Paul G Farnham, Timothy J Sullivan, Berry Bennett, Peter R Kerndt, et al. 2010. “Cost-Effectiveness of Pooled Nucleic Acid Amplification Testing for Acute HIV Infection after Third-Generation HIV Antibody Screening and Rapid Testing in the United States: A Comparison of Three Public Health Settings.” PLoS Medicine 7 (9): e1000342. ICF International, Lisa Robinson, and James Hammitt. 2009. Final Report: Sub-Saharan Africa Refinery Project Health Study: Volume 1-B Appendices. Washington, DC. http://www.unep.org/transport/pcfv/PDF/Health_Study_Volume_I-B_6-4-09.pdf. Johnson, Johnny. 2009. “Absenteeism Trends in South African Companies.” Human Capital Review: 1-2. Kalichman, Seth C, Leickness C Simbayi, Michelle Kaufman, Demetria Cain, and Sean Jooste. 2007. “Alcohol Use and Sexual Risks for HIV/AIDS in Sub-Saharan Africa: Systematic 26 Review of Empirical Findings.” Prevention Science β―: The Official Journal of the Society for Prevention Research 8 (2): 141–51. LΚΌEngle, Kelly L, Peter Mwarogo, Nzioki Kingola, William Sinkele, and Debra H Weiner. 2014. “A Randomized Controlled Trial of a Brief Intervention to Reduce Alcohol Use among Female Sex Workers in Mombasa, Kenya.” Journal of Acquired Immune Deficiency Syndromes 67 (4): 446–53. Matzopoulos, Richard Gregory, Sarah Truen, Brett Bowman, and Joanne Corrigall. 2014. “The Cost of Harmful Alcohol Use in South Africa.” South African Medical Journal 104 (2): 127–32. Montez, David. 2010. “Mobile Communications in Zambia.” InterMedia Survey Institute. http://www.audiencescapes.org/intermedia-africa-research-reports-mobile-communicationszambia-phone-sharing-banking-barriers-access-demand-side-survey. Moore, Mark A., Anthony E. Boardman, Aidan R. Vining, David L. Weimer, and David H. Greenberg. 2004. “‘Just Give Me a Number!’ Practical Values for the Social Discount Rate.” Journal of Policy Analysis and Management 23 (4): 789–812. Mortimer, Duncan, and Leonie Segal. 2005. “Economic Evaluation of Interventions for Problem Drinking and Alcohol Dependence: Cost per QALY Estimates.” Alcohol and Alcoholism (Oxford, Oxfordshire) 40 (6): 549–55. National Aids Council. 2014. Zambia Country Report: Monitoring the Declaration of Commitment on HIV and AIDS and the Universal Access. Lusaka. http://www.unaids.org/sites/default/files/country/documents/ZMB_narrative_report_2014.p df. Nicholson, Sean, Mark V Pauly, Daniel Polsky, Claire Sharda, Helena Szrek, and Marc L Berger. 2006. “Measuring the Effects of Work Loss on Productivity with Team Production.” Health Economics 15 (2): 111–23. Norris, Lexi, Leslie Swartz, and Mark Tomlinson. 2013. “Mobile Phone Technology for Improved Mental Health Care in South Africa: Possibilities and Challenges.” South African Journal of Psychology 43 (3): 379–88. Nyonator, Frank K, J Koku Awoonor-Williams, James F Phillips, Tanya C Jones, and Robert a Miller. 2005. “The Ghana Community-Based Health Planning and Services Initiative for Scaling up Service Delivery Innovation.” Health Policy and Planning 20 (1): 25–34. Paltzer, Jason Timothy. 2014. “Substance Use Screening, Brief Intervention, and Referral to Treatment Among Working-Age Medicaid Patients in Wisconsin: Impacts on Healthcare Utilization.” University of Wisconsin-Madison. Paltzer, Jason. 2009. "A Preliminary Study of Alcohol Perceptions in Zambia." Working paper. Papas, Rebecca K., John E. Sidle, Steve Martino, Joyce B. Baliddawa, Rogers Songole, Omolo E. Otieno, Benson N. Gakinya, et al. 2010. “Systematic Cultural Adaptation of CognitiveBehavioral Therapy to Reduce Alcohol Use among HIV-Infected Outpatients in Western Kenya.” AIDS and Behavior 14 (3): 669–78. Pauly, Mark V., Sean Nicholson, Daniel Polsky, Marc L. Berger, and Claire Sharda. 2008. “Valuing Reductions in on-the-Job Illness: ‘Presenteeism’ from Managerial and Economic Perspectives.” Health Economics 17 (4): 469–85. 27 Pidd, Kenneth J., Jesia G. Berry, Ann M. Roche, and James E. Harrison. 2006. “Estimating the Cost of Alcohol-Related Absenteeism in the Australian Workforce: The Importance of Consumption Patterns.” Medical Journal of Australia 185 (11-12): 637–41. Quanbeck, Andrew, Katharine Lang, Kohei Enami, and Richard L. Brown. 2010. “A Cost-Benefit Analysis of Wisconsin’s Screening, Brief Intervention, and Referral to Treatment Program: Adding the Employer's Perspective.” Wisconsin Medical Journal 109 (1): 9–14. Rehm, Jürgen, Colin Mathers, Svetlana Popova, Montarat Thavorncharoensap, Yot Teerawattananon, and Jayadeep Patra. 2009. “Global Burden of Disease and Injury and Economic Cost Attributable to Alcohol Use and Alcohol-Use Disorders.” Lancet 373 (9682): 2223–33. Russell, Louise B. 2009. “Completing Costs: Patients’ Time.” Medical Care 47 (7): 89–93. Schatz, Joseph J. 2008. “Zambia’s Health-Worker Crisis.” The Lancet 371 (9613): 638–39. Schoen Cathy, Robin Osborn, David Squires, and Michelle M. Doty. 2013. "Access, Affordability, and Insurance Complexity are Often Worse in the United States Compared to Ten Other Countries." Health Affairs 32(12): 2205-2215. Solberg, Leif I, Michael V Maciosek, and Nichol M Edwards. 2008. “Primary Care Intervention to Reduce Alcohol Misuse Ranking Its Health Impact and Cost Effectiveness.” American Journal of Preventative Medicine 34 (2): 143–52. State Council on Alcohol and Other Drug Abuse (SCAODA). 2013. Screening Brief Intervention and Referral to Treatment (SBIRT) Report to the State Council on Alcohol and Other Drug Abuse. Swider, Susan M, and D Ph. 2002. “Outcome Effectiveness of Community Health Workers β―: An Integrative Literature Review.” Public Health Nursing 19 (1). Thavorncharoensap, Montarat, Yot Teerawattananon, Jomkwan Yothasamut, Chanida Lertpitakpong, Khannika Thitiboonsuwan, Prapag Neramitpitagkul, and Usa Chaikledkaew. 2010. “The Economic Costs of Alcohol Consumption in Thailand, 2006.” BMC Public Health 10 (323). UN AIDS. 2014. “Zambia HIV and AIDS Estimates (2013).” Countries. http://www.unaids.org/en/regionscountries/countries/zambia. U.S. Department of Transportation. "Guidance on Treatment of the Economic Value of a Statistical Life in U.S. Department of Transportation Analyses." Undated memorandum. Accessed 11/8/2014. http://www.dot.gov/sites/dot.dev/files/docs/VSL%20Guidance%202013.pdf Viscusi, W. Kip, and Joseph E. Aldy. 2003. “The Value of a Statistical Life: A Critical Review of Market Estimates Throughout the World.” Journal of Risk and Uncertainty 27 (1). Kluwer Academic Publishers: 5–76. Wisconsin Department of Health Services (WDHS). 2012. Wisconsin Epidemiological Profile on Alcohol and Other Drug Use, 2012. Office of Health Informatics, Division of Public Health, in consultation with the Division of Mental Health and Substance Abuse Services and the University of Wisconsin Population Health Institute. September 2012. Wisconsin Initiative to Promote Healthy Lifestyles (WIPHL). 2012. Final Progress Report to the Substance Abuse and Mental Health Services Administration SBIRT Grant TI-18309. 28 World Bank. 2014. “World Databank”. October 14. http://data.worldbank.org/indicator/FP.CPI.TOTL.ZG World Health Organization (WHO). 2013. Global Status Report on Road Safety 2013 Supporting a Decade of Action. Geneva, Switzerland. http://www.who.int/violence_injury_prevention/road_safety_status/2013/en/. ———.2014. Global Status on Health and Alcohol Report. Geneva, Switzerland. http://www.who.int/substance_abuse/publications/global_alcohol_report/msb_gsr_2014_1.p df?ua=1. Yende, Pearl Monica. 2005. “Utilising Employee Assistance Programmes to Reduce Absenteeism in the Workplace.” Zambia’s Ministry of Health. 2011. National Health Strategic Plan 2011-2015. Zambia’s Ministry of Health and World Health Organization Country Office Zambia. 2008. Prevalence Rates of the Common Non- Communicable Diseases and Their Risk Factors in Lusaka District , Zambia Prevalence Rates of the Common Non-Communicable Diseases and Their Risk Factors in Lusaka District , Zambia 2008. Lusaka. Zambian Road Safety Trust. 2014. “Prevalence of Road Accidents in Zambia.” Accessed October 18. http://zambianroadsafety.org/press-release-prevalence-of-road-accidents-in-zambia/. Zurovac, Dejan, Bruce a Larson, Raymond K Sudoi, and Robert W Snow. 2012. “Costs and CostEffectiveness of a Mobile Phone Text-Message Reminder Programmes to Improve Health Workers’ Adherence to Malaria Guidelines in Kenya.” PloS One 7 (12): 1–6. 29 APPENDICES APPENDIX A: ZAMBIAN DEMOGRAPHICS The Central Statistics Office’s 2008 Labor Force Survey Report offers the most recent demographic data. Our demographic parameters are directly taken or calculated from this report. It is our assumption that these data are still relevant in 2014 with only minor changes. This assumption is confirmed by the CSO’s preliminary 2012 labor force findings. The preliminary 2012 Labor Force Survey Report is available; however, it lacks the detail of the report in 2008. Based on the time of release of the 2008 report, the full 2012 report may be released in the 2014 to 2016 timeframe. The following demographic information for the pilot village (Mwembeshi) and Central Province is incorporated into our Monte Carlo simulation. FIGURE 4: DEMOGRAPHIC CHARACTERISTICS Source: Authors 30 Table 9: Demographic Characteristics Sex Male Female Geographic Location Rural Urban Economic Status Economically Active Economically Inactive Employment Status Employed Unemployed Sector of Employment Informal Formal Source: CSO 2008 Economic Status is divided into two categories (1) economically active and (2) economically inactive. Economically active denotes an individual that is either employed or unemployed (CSO 2008). Economically inactive denotes an individual that is outside the labor force. The CSO includes full-time students, full-time homemakers, retired individuals, beggars, etc. in the economically inactive population (CSO 2008). Table 10: Demographic Data Specific to Pilot Village Table 10a: Sex Male (%) 49 Female (%) 51 Source: CSO 2008 Table 10b: Geographic Location Rural (%) 100 Source: CSO 2008 Table 10c: Economic Status: Economically Active (includes employed and unemployed) Rural Male (%) Female (%) 82.3 80 31 Table 10d: Employment Status: Unemployed Rural Male (%) Female (%) 4.0 3.4 Source: CSO 2008 Table 10e: Mwembeshi Sector of Employment, Rural Male (%) Female (%) Formal 6.24 2.19 Informal 93.76 97.81 Source: CSO 2008 Table 11: Demographic Data Specific to Central Province Table 11a: Sex Male (%) 49 Female (%) 51 Source: CSO 2008, 2013 Table 11b: Geographic Location Rural (%) 76 Urban (%) 24 Source: CSO 2008, 2013 Table 11c: Economic Status: Economically Active Male (%) Female (%) Rural 82.3 80 Urban 73 52.9 Source: CSO 2008, 2013 32 Table 11d: Employment Status: Unemployed Male (%) Female (%) Rural 4.0 3.4 Urban 10.8 11.7 Source: CSO 2008, 2013 Table 11e: Sector of Employment, Rural Male (%) Female (%) Formal 6.24 2.19 Informal 93.76 97.81 Source: CSO 2008, 2013 Table 11f: Sector of Employment, Urban Male (%) Female (%) Formal 35.77 20.26 Informal 64.23 79.74 Source: CSO 2008, 2013 33 APPENDIX B: ADJUSTING FOR THE EXCHANGE RATE Converting our estimates to United States Dollars (USD) required one to two calculations depending on the estimate. If the estimate was in ZWK, the Zambian currency prior to the 2012 debasement, we divided the estimate by 1,000 to equate the ZWK with the new currency ZWM. One thousand (1,000) ZWK is equal to one (1) ZWM. To convert ZWM to USD, multiply the ZWM estimate by .1577. This is the official exchange as of November 19, 2014. We acquire this exchange rate from the company XE. 1 1 XE can be accessed at the following link: http://www.xe.com/currencyconverter/convert/?Amount=1&From=ZMW&To=USD 34 APPENDIX C: ADJUSTING FOR INFLATION Throughout the cost-benefit analysis it was necessary to translate estimates across time to 2014 kwachas and dollars. The World Bank offers two measures of inflation, which include: the Gross Domestic Product (GDP) deflator and the Consumer Price Index (CPI). To be consistent, all estimates of costs and benefits that need be adjusted for inflation are translated using inflation in consumer prices (CPI). Both measures are reported as the annualized increase in prices over one calendar year. If able to determine the wage inflation rate within each sector of the economy (i.e. rural informal, rural formal, urban informal, urban formal) a more accurate wage figure can be incorporated into the model. Below is a table representing the various inflation rates for the years 2009 to 2013. Table 12: Zambia Inflation Rates Consumer Prices Year Rate (%) 2009 13.4 2010 8.5 2011 6.4 2012 6.6 2013 7.0 Source: World Bank 2014 35 APPENDIX D: CALCULATION OF AVERAGE WAGES IN ZAMBIA Using the 2008 labor report issued by the CSO, we obtained various wage rates based on gender, geographic location, and sector of employment. We adjusted the 2008 average monthly wage rates to 2014 average monthly wage rates by adjusting for inflation for years 2009-2013 (See Appendix E). Using the consumer price index, we obtained values in 2014 dollars. The 2008 inflation rate was not included in our adjustment because the CSO conducted the 2008 labor survey in the November/ December timeframe of year 2008. We use these eight wage rates in the calculation of our social network costs, patient opportunity costs, and productivity benefits. Following our inflation adjustment, the ZMK was debased in 2012 with the introduction of a new currency ZMW where 1,000 ZMK is equal to one (1) ZMW. We converted ZMW to USD using the exchange rate as of November 19, 2014; with .1577 USD equal to 1 ZMW. To conclude with our final wage estimates, we multiplied the average monthly 2014 USD wage by 12 to calculate the average annual 2014 USD wage. The following wage calculation is applied to all eight wages displayed in the two tables below. Inflation adjusted average 2014 ZMK wage = (average monthly 2008 wage ZMK) * (respective inflation rates for the years 2009 to 2013) Adjusted for new currency 2014 ZMW wage = (inflation adjusted 2014 ZMK wage) / (1000) Average monthly 2014 wage = (adjusted for new currency 2014 ZMW wage) * (.15577) (USD/ZMW) Average annual 2014 wage (USD) = (average monthly 2014 wage) * (12 months) These eight wages are displayed in the tables below. 36 Table 13: Female Wage Rates Female Wage Rate Category Average monthly 2008 wage (ZMK) Inflation adjusted average monthly 2014 wage (ZMK) Adjusted for new currency ZMW Average monthly 2014 wage (USD) (1000 ZKW = 1 ZWM) (.1577 USD = 1 ZMW) Average annual 2014 wage (USD) Formal & Rural 923,481 1,270,937 1,271 200 2405 Formal & Urban 1,866,801 2,787,555 2,788 440 5275 Informal & Rural 232,814 347,644 348 55 658 Informal & Urban 766,338 1,144,315 1,144 180 2166 Inflation adjusted average monthly 2014 wage (ZMK) Adjusted for new currency ZMW Average monthly 2014 wage (USD) Average annual 2014 wage (USD) (1000 ZKW = 1 ZWM) (.1577 USD = 1 ZMW) Source: CSO (2008); Authors Table 14: Male Wage Rates Male Wage Rate Category Average monthly 2008 wage (ZMK) Formal & Rural 1,052,770 1,572,023 1,572 247 2,974 Formal & Urban 2,638,043 3,939,193 3,939 621 7,454 Informal & Rural 358,475 535,284 535 84 1,012 Informal & Urban 1,129,320 1,686,329 1,686 265 3,191 Source: CSO (2008); Authors 37 APPENDIX E: CALCULATION OF ABSENTEEISM COSTS Absenteeism costs under current policy are accrued as benefits following the successful treatment of a drinker. Absenteeism is defined as the days absent from work and in our case the days absent from work attributed to alcohol consumption (Baumberg 2010). These costs affect the individual drinker, the employer of the individual drinker, and the colleagues of the individual drinker. Costs vary primarily by the drinking state and wage of each worker. Our estimates are calculated per individual. We estimate benefits in terms of an individual moving from a drinking state to an abstainer state with a gradual reduction in benefits overtime. Pidd et al. (2006) estimates that there is a positive correlation in sick-absenteeism in the workplace and alcoholic consumption and frequency. Pidd et al. (2006) uses a comprehensive survey in Australia on sick-related absenteeism that divides drinkers into various categories similar to SBIRT. The advantage of using the Pidd et al. (2006) analysis is the ability to assign absenteeism based on drinking state. Adjusting these values for SBIRT we were able to obtain an average number of days per drinking category. These values are displayed in the table below. Table 15: Means days absent attributed to sickness by drinking category in Australia Drinking Category Average Days Off Abstainers 5.51 Low Risk 5.84 At-risk 7.04 Harmful 7.05 Likely dependent 9.95 Source: Pidd et al. (2006); Authors 38 The weighted average based on the population statistics provided in Pidd et al. (2006) is 6.6 days absent attributed to sickness. Under the assumption of 250 working days, the Australian sick-absenteeism rate is estimated at approximately 2.6 percent. To adjust the overall absenteeism days by drinking category to Zambia, we use a 2011 South African sick-related absenteeism rate of 3.40 percent (Alexander Forbes Health Management Solutions 2013). We assume that the South African rate is a reasonable approximate estimate for Zambia. Absenteeism can depend upon cultural elements and given the high prevalence of problem drinking in Zambia relative to Australia and the United States, it is plausible that these estimates are conservative. We recommend that if a Zambian work-related absenteeism rate is surveyed and discovered for both the formal and informal sector that our estimates be revised accordingly. We adjusted days absent for each drinking state by a ratio factor of 1.292 (.0340/.0263) to account for the difference in overall absenteeism between South Africa and Australia. Upper and lower bounds are obtained by applying an absenteeism range specific to South Africa provided by Johnson (2009). The overall absenteeism range of 3.5 percent and 6.0 percent is translated into an overall sick-absenteeism range of 2.51 percent and 4.29 percent using ratios between upper and lower bounds from the average value. Average, upper bounds, and lower bounds are provided in the table below. 39 Table 16: Means days absent attributed to sickness by drinking category in Zambia (transferred from South Africa context) Drinking State Lower Bound Average Upper Bound Abstainers 5.24 7.12 8.99 Low Risk 5.56 7.54 9.52 At-risk 6.70 9.10 11.49 Harmful 6.71 9.10 11.50 Likely dependent 9.47 12.85 16.22 Source: Authors Consequently, to determine absenteeism-related to alcohol, we subtract the mean sickrelated absenteeism days for the abstainers from each drinking class to obtain the following estimates of alcohol-related absenteeism days. Pidd et al. (2006) employs this methodology. Table 17: Days absent associated with alcohol consumption per year Drinking State Lower Bound Average Upper Bound Abstainers 0.00 0.00 0.00 Low Risk 0.31 0.43 0.54 At-risk 1.46 1.98 2.50 Harmful 1.46 1.98 2.51 Likely dependent 4.22 5.73 7.24 Source: Authors; Pidd et. al (2006) methodology We convert days missed into “percent of working year absent attributed to alcohol consumption,” to account for the wide discrepancies in work hours by sector, geographical location and sex rather than using a “number of days absent” measure. We assume 250 working days (assumed in both Pidd et al. 2006 and Johnson 2009) to calculate the percent lost and 40 calculate the following percentages. Our Monte Carlo simulation assumes a triangular distribution. Table 18: Percent of working year absent attributed to alcohol consumption Drinking State Lower Bound (%) Average (%) Upper Bound (%) Abstainers 0.00 0.00 0.00 Low Risk 0.12 0.17 0.22 At-risk 0.58 0.79 1.00 Harmful 0.59 0.79 1.00 Likely dependent 1.69 2.29 2.90 Source: Authors The utility of using a “percentage of the year absent,” under the assumption that these percentages remain constant across working classes, is that we are able to apply the percentage to different types of workers that may vary in days and hours worked. If a comprehensive survey in Zambia collected the days absent missed by geographic location, sex, and sector of work would greatly advance our ability to provide a more detailed Zambian-specific estimate. Furthermore, Yende discusses absenteeism among South African males citing that each male alcoholic interviewed missed 86 working days a year due to absence (Yende 2005). However, because of a lack of data, differences between male and female, and other concerns we deployed only the conservative estimates based off the Australian survey data. Absenteeism costs will vary on the drinking state, wage, and sector of employment for a given individual. Each baseline absenteeism cost under current policy is calculated the following: Baseline Absenteeism Cost = (wage) * (percent of working year absent attributed to alcohol) 41 If the individual works in the formal sector economy, we assume there is an average spillover multiplier of 1.1 with a uniform distribution with lower and upper bounds of 1.0 to 1.2, respectively. This spillover multiplier accounts for the increased costs to the employer or fellow colleagues due to the absence of the drinker. A spillover of 1.1 is used for Zambia due to the high concentration of low-wage, low-skilled and easily replaceable labor (Quanbeck et al. 2010; Nicholson et al. 2006). We estimated total absenteeism cost savings by sector as follows (Quanbeck et al. 2010; Baumberg 2010): If formal sector: Absenteeism Cost = (absenteeism multiplier) * (wage) * (percent of working year absent attributed to alcohol) If informal sector: Absenteeism Cost = (1.0) * (wage) * (percent of working year absent attributed to alcohol) The overall range for women varies between $0.83 and $183.30. Estimates vary with drinking state, wage that is conditional on location and sector of employment, as well as, the absenteeism multiplier. As stated, these costs are incorporated into our model accrued as benefits following successful treatment. The table below summarizes the range of potential estimates within categories. 42 Table 19: Female Absenteeism Cost Estimates (USD) Drinking State Lower Bound ($) Average ($) Upper Bound ($) Low Risk 3.02 4.51 6.22 At-risk 14.03 20.94 28.86 Harmful 14.06 20.99 28.93 Likely Dependent 40.63 60.65 83.57 Low Risk 6.63 9.89 13.63 At-risk 30.77 45.93 63.29 Harmful 30.85 46.05 63.45 Likely Dependent 89.10 133.02 183.30 Low Risk 0.83 1.12 1.42 At-risk 3.84 5.21 6.58 Harmful 3.85 5.22 6.59 Likely Dependent 11.11 15.08 19.05 Low Risk 2.72 3.69 4.66 At-risk 12.63 17.14 21.65 Harmful 12.66 17.18 21.71 Likely Dependent 36.58 49.64 62.70 Formal & Rural Formal & Urban Informal Rural Informal Urban Source: Authors The overall range for men varies between $1.27 and $259.03. The table below summarizes the range of potential estimates within categories. 43 Table 20: Male Absenteeism Cost Estimates (USD) Drinking State Lower Bound ($) Average ($) Upper Bound ($) Low Risk 3.74 5.58 7.69 At-risk 17.35 25.90 35.69 Harmful 17.40 25.97 35.78 Likely Dependent 50.25 75.01 103.37 Low Risk 9.36 13.98 19.26 At-risk 43.48 64.90 89.44 Harmful 43.59 65.07 89.67 Likely Dependent 125.91 187.97 259.03 Low Risk 1.27 1.73 2.18 At-risk 5.91 8.02 10.13 Harmful 5.92 8.04 10.15 Likely Dependent 17.11 23.22 29.33 Low Risk 4.01 5.44 6.87 At-risk 18.61 25.26 31.91 Harmful 18.66 25.32 31.99 Likely Dependent 53.90 73.15 92.41 Formal & Rural Formal & Urban Informal Rural Informal Urban Source: Authors The distribution of benefits will vary with the different demographic parameters specified for the pilot phase and Central Province. 44 APPENDIX F: CALCULATION OF PRESENTEEISM COSTS Presenteeism costs under currency policy are accrued as benefits following the successful treatment of a drinker. Presenteeism measures the reduced productivity at work for a given drinker, which may be attributed to hangovers, on-the-job drinking, reduced work effectiveness resulting from alcohol-induced health disabilities, etc. (Baumberg 2010). Costs depend on the drinking state of each respective patient. Our estimates are calculated per individual. We estimate benefits in terms of an individual moving from a drinking state to an abstainer state with a gradual reduction in benefits overtime. Because of lack of data, cultural sensitivities, and existing methodological controversies, it is necessary to state several assumptions. We assume that absenteeism rates and presenteeism rates are correlated. More specifically, we assume that absenteeism and presenteeism occur at comparable levels and increase jointly as an individual increases their level of drinking (Anderson 2010; Pidd et al. 2006). Additionally, we assume that the rate of absenteeism and presenteeism between drinking states remain constant across countries and time allowing the use of western data to adjust in a Zambian context. Furthermore, we assume the rates of change (for absenteeism and presenteeism) between drinking classes do not change. We calculate presenteeism days attributed to alcohol based on a cost-benefit analysis in the United States and adjust for Zambia. Arora et al. (2011) estimates a range of days that a likely alcohol dependent experiences presenteeism: 8.7 to 22.51 days. Based on the assumption of a positive correlation between absenteeism and presenteeism, we multiply this range by a factor equal to the ratio of the 2011 US sick-related absenteeism rate and the South African sickrelated absenteeism rate. This factor equals 1.545 calculated by dividing the 2011 South African sick-related absenteeism rate (3.4 percent) by the 2011 US sick-related absenteeism rate (2.2 45 percent) (Alexander Forbes Health Management Solutions 2011; BLS 2011). Adjusting to South Africa, which we assume to be a plausible proxy to the Zambian context, we estimate a range of 13.45 to 34.79. The range is adjusted and held constant from the US context. Table 21: Sick-related Absenteeism Rate Country Sick-related absenteeism rate (2011) (%) United States 2.2 South Africa (Zambia) 3.4 Source: BLS (2013); Alexander Forbes Health Management Solutions (2011) Table 22: Adjusting US Presenteeism Days to Zambia the SBIRT Likely Dependent drinking Country Lower Bound Average Upper Bound US Likely Dependent 8.70 15.61 22.51 Zambia Likely Dependent 13.44 24.12 34.79 Source: Authors To adjust presenteeism days to the other SBIRT drinking states, we use the likely dependent drinker as the base case and adjust the estimate downward using the ratio that absenteeism varies between drinking categories. We assume that absenteeism and presenteeism ratios do not vary across countries and offers a plausible adjustment factor relating absenteeism and presenteeism across different drinking states. We estimate the following: 46 Table 23: Ratio of days absent between drinking states Drinking State Average days absent attributed to alcohol Ratio between drinking states* Calculation Abstainers 0.000 0.000 Abstainer / Low Risk Low Risk 0.426 0.215 Low Risk / At-risk At-risk 1.979 0.997 At-risk / Harmful Harmful 1.984 0.346 Harmful / Likely Dependent Likely dependent 5.731 *Assumption: Ratios hold true when applied to absenteeism and presenteeism across countries. Source: Authors Table 24: Days presenteeism occurs associated with alcohol consumption per year Drinking State Lower Bound Average Upper Bound Abstainers 0.00 0.00 0.00 Low Risk 1.00 1.79 2.58 At-risk 4.64 8.33 12.01 Harmful 4.66 8.35 12.04 Likely Dependent 13.45 24.12 34.79 Source: Authors Similar to absenteeism, we state presenteeism as a percentage of working days in order to apply across large wage differentials. Our Monte Carlo simulation assumes a triangular distribution. The following table exhibits the percentages calculated. 47 Table 25: Percent of working year presenteeism occurs attributed to alcohol consumption Drinking State Lower Bound (%) Average (%) Upper Bound (%) Abstainers 0.00 0.00 0.00 Low Risk 0.40 0.72 1.04 At-risk 1.86 3.33 4.81 Harmful 1.86 3.34 4.82 Likely Dependent 5.38 9.65 13.92 Source: Authors Presenteeism costs depend on the drinking state, wage, and sector of employment for a given individual. Furthermore, wage is further divided into sex, geographic location, and sector of employment. Regarding the sector of employment, we assume that only individuals in the formal sector of the economy will accrue presenteeism cost savings that depend on an impairment factor. We use an impairment factor of 10 percent with a uniform distribution with lower and upper bounds of zero and twenty respectively. We justify using a low impairment factor for the overall low skill and relatively easy replaceable labor market characteristics (Pauly et al. 2008; Quanbeck et al. 2010). An impairment factor measures the degree of reduced productivity that will occur in the workplace when an individual comes hung over, drunk, etc. Each presenteeism cost under currency policy is estimated the following (Quanbeck et al. 2010; Thavorncharoensap et al. 2010): If formal sector: Presenteeism Costs = (impairment factor) * (wage) * (percent of working year presenteeism occurs attributed to alcohol) If informal sector: 48 Presenteeism Cost = zero (0) The overall range for women varies between $0.00 and $146.81. Estimates vary with drinking state, wage that is conditional on location and sector of employment, as well as, the presenteeism impairment factor. As stated, these costs are incorporated into our model as benefits following successful treatment. The table below summarizes the range of potential estimates within categories. Table 26: Female Presenteeism Cost Estimates (USD) Drinking State Lower Bound Average ($) ($) Upper Bound ($) Low Risk 0.00 1.73 4.98 At-risk 0.00 8.01 23.11 Harmful 0.00 8.03 23.17 Likely Dependent 0.00 23.20 66.94 Low Risk 0.00 3.78 10.92 At-risk 0.00 17.57 50.69 Harmful 0.00 17.62 50.82 Likely Dependent 0.00 50.89 146.81 Low Risk 0.00 0.00 0.00 At-risk 0.00 0.00 0.00 Harmful 0.00 0.00 0.00 Likely Dependent 0.00 0.00 0.00 Formal & Rural Formal & Urban Informal Rural 49 Low Risk 0.00 0.00 0.00 At-risk 0.00 0.00 0.00 Harmful 0.00 0.00 0.00 Likely Dependent 0.00 0.00 0.00 Informal Urban Source: Authors The overall range for men varies between $0.00 and $207.46. Estimates vary with drinking state, wage that is conditional on location and sector of employment, as well as, the presenteeism impairment factor. The table below summarizes the range of potential estimates within categories. Table 27: Male Presenteeism Cost Estimates (USD) Drinking State Lower Bound Average ($) ($) Upper Bound ($) Low Risk 0.00 2.13 6.16 At-risk 0.00 9.91 28.59 Harmful 0.00 9.93 28.66 Likely Dependent 0.00 28.70 82.79 Low Risk 0.00 5.35 15.43 At-risk 0.00 24.83 71.63 Harmful 0.00 24.89 71.82 Likely Dependent 0.00 71.91 207.46 Formal & Rural Formal & Urban 50 Low Risk 0.00 0.00 0.00 At-risk 0.00 0.00 0.00 Harmful 0.00 0.00 0.00 Likely Dependent 0.00 0.00 0.00 Low Risk 0.00 0.00 0.00 At-risk 0.00 0.00 0.00 Harmful 0.00 0.00 0.00 Likely Dependent 0.00 0.00 0.00 Informal Rural Informal Urban Source: Authors The distribution of benefits will vary with the different demographic parameters specified for the pilot phase and Central Province. 51 APPENDIX G: CALCULATION OF PATIENT OPPORTUNITY COST The incorporation of patient costs into cost-benefit analyses of health interventions appears mixed in the literature. Some analyses do not include any costs that accrue only to the patient, but instead focus on provider costs (Bray et al. 2012). Another set of studies argue that ignoring these costs underestimates disease burden and biases results, stating further that even if the amount of time appears negligible, when aggregated it could significantly impact findings (Russell 2009). In an effort to guard against the possibility of overstating the benefits of SBIRT in Zambia, especially in view of the significant uncertainty in the acceptance and success rates of the SBIRT program, we have chosen the more conservative approach of including patient costs in our analysis. The opportunity costs for patients takes into account alternative uses of the time spent receiving health care services. The time spent receiving services is a summation of expected time allocated to the brief screen, brief intervention (in cases of positive screens), and time spent receiving treatment as well as time waiting for and traveling to and from treatment at the nearest referral center (in cases of likely dependency). Other studies that have incorporated patient time costs have used the average wage rate as a proxy for the patients’ opportunity cost (Fleming et al. 2000). This is in agreement with economic literature on opportunity cost valuations, which posits that the wage rate is the price of labor as well as the price of leisure since an hour’s wage is the price of consuming an additional hour of leisure (Russell 2009). The wage rate used to determine the patient opportunity cost in this cost-benefit analysis is the same as that used in the absenteeism and presenteeism calculations. This accounts for whether the patient is economically active or inactive, employed or unemployed, in a rural or urban area, in a formal or informal sector of the economy, and 52 whether the patient is male or female. Based on distributions of these characteristics in the population, an observation in the Monte Carlo simulation is assigned a corresponding wage. One distinction is made, however, with respect to the likely dependent drinking category. For this classification of patients, to take account of the additional burden of traveling to a clinic to receive care and the significant time commitment, we factored in a travel cost and a retention rate. The travel cost used is an estimate of the fuel cost, while the retention rate accounts for the fact that not all who are referred to the clinic actually go to receive treatment. This rate is based on a systematic review of adult patient attrition from ART programs in service delivery settings in sub-Saharan Africa, which found that on average 70 percent remain in treatment after 24 months (Fox and Rosen 2010). A travel cost is not employed for those in the other drinking categories. All screening and treatment takes place in the patient’s home for these other cases, and therefore, the patient requires no travel. The simulated wage is then converted into the proper time unit and multiplied by the estimated range of time spent receiving treatment. For all in the sample, this was 15 minutes for the brief screen, with a range of 10 to 20 minutes, and moderated by the participation rate. For those with at-risk drinking, the estimate is 45 to 65 minutes, with an estimated 15 minutes for the brief screen, 20 minutes for the full screen, and 1 to 3 counseling sessions at 15 minutes each. For harmful drinkers, this time allotment becomes 50 to 80 minutes, with 15 minutes for the brief screen, 20 minutes for the full screen, and a range of 3 to 6 intervention sessions at 15 minutes each. Lastly, for those with likely dependency, in view of high uncertainty in the travel costs, time spent waiting to receive treatment, and the duration of treatment, a large range was used. In the simulation, however, those who were classified as likely dependent but did not accept treatment were only assigned the cost for the brief and full screens. 53 One nearby clinic (SHARPZ clinic) offers a five-day intensive alcohol treatment. With this in mind, we used five days as our point estimate for those likely dependent, again, using the daily wage as the proxy for opportunity cost. A range of plus or minus two days was then incorporated into our sensitivity analysis to account for the large degree of uncertainty with this aspect of the opportunity cost estimate. The travel cost ranged from 0, if the travel could be incorporated into another trip to Lusaka, to a high of 25 dollars, if the travel was undertaken by an individual in a personal vehicle. Carpooling or use of public transportation would decrease the cost, but would remain in the above range. A simulated acceptance of treatment was then used based on a distribution using the retention rate mentioned above, with two levels of 0 and 1. Thus, if 0, no opportunity cost apart from the brief and full screens were accounted for, while for a 1, the complete list of costs were included. 54 APPENDIX H: CALCULATION OF SOCIAL NETWORK COSTS The social network costs are calculated as a loss in employment opportunities from no longer drinking with a cohort of friends. This measure is included in response to evaluations of the Zambian drinking culture. These observations stress the centrality of drinking in some social circles, and how changing an individual’s drinking habits would likely result in a disruption of his or her existing social network (Haworth 2004). Thus, with this estimation we are assuming that the individual will be excluded from his or her previous social network as a result of the changed alcohol consumption and that no new network is formed during the time period of the program. To estimate this impact, we use information gathered from an International Labour Office study on the Zambian labor market (ILO 2013). In this research it was found that roughly 31 percent of youth found their employment through friends, relatives, and acquaintances. As some members of this social network would not change as a result of decreased drinking, we use a range from zero to 31 percent. This percentage of employment gained through referrals was then multiplied by a composite of the annual wage and the average length of transition to new employment. The transition statistic used in the calculations showed an average length of transition of 1.6 months for men and 2.9 months for women (ILO 2013). The composite wage was separate for men and women, and was weighted by the distribution of rural and urban areas and the distribution of formal and informal employment specific to the pilot area and Central Province. This is only included in the model for successfully treated drinkers that are unemployed. 55 APPENDIX I: CALCULATION OF HEALTH CARE SYSTEM COST The health care system costs are the change in the burden on nearby health care centers as a result of the SBIRT program. A Wisconsin study found that implementing SBIRT annually increases outpatient days by 4.5 per person, while decreasing in-patient days by 2 (Paltzer 2014). This is consistent with other studies that find SBIRT reduces health costs by reducing the number of emergency department visits and hospitalizations (Agerwala and McCance-Katz 2012; Gentilello et al. 2005). Moreover, the increase in outpatient days found in the Wisconsin study can also be used to factor in the increased outpatient days resulting from the referrals to treatment, and other possible visits resulting from increased awareness of health consequences of drinking. To then translate this into the Zambian context, we accounted for differences in the propensity to use health care services and differences in health care costs. Based on survey data from the two countries we calculated a percent difference in use of health care facilities when ill: Zambians are 35 percent less likely to seek care (Schoen et al. 2013; Hjortsberg 2003). This percent difference was then used to adjust the expected changes in inpatient and outpatient service utilization resulting from SBIRT. The second translation mechanism used average cost estimates of outpatient and inpatient days from a Zambian study. The study found inpatient costs to be $18 per bed day and outpatient costs to be $3 per visit (Chola and Robberstad 2009). These statistics were then converted to 2014 USD, which resulted in estimates of $19.97 per bed day for inpatient services and $3.33 per visit for outpatient services. Using these cost estimates and the adjusted expected utilization changes gave an estimated $16.30 in annual cost savings per person who participates in SBIRT. 56 APPENDIX J: CALCULATION OF AVOIDED COSTS OF CRIME AND ROAD TRAFFIC ACCIDENTS AVOIDED COSTS OF CRIME To approximate the cost of alcohol-related crime, we used estimates of the cost of harmful alcohol use in South Africa; the aggregate cost of crime was $3.27 billion or 7.8 percent of GDP. Evidence shows that in low- to middle-income countries in Latin America, with high crime levels, crime costs range from 5 to 15 percent of GDP (Matzopoulos et al. 2014). Matzopoulos et al. (2014) apply an alcohol-attributable fraction (AAF) of crime of 0.25 and assume that 75 percent of costs of crime are relevant to alcohol use, which include response of crime, consequences of crime such as damages to victims, anticipation of crime, and nonfinancial welfare costs which includes the emotional costs associated with alcohol-related crime. Under these assumptions, alcohol-attributable crime costs totaled $4.14 billion (1.45 percent of GDP) in South Africa. In another study Boucher et al. (2006) estimate the economic costs of alcohol consumption in the United States. Victim costs, criminal justice system costs, and productivity loss represented $73.327 billion in 2006 (0.53 percent of GDP). Using Zambia’s 2013 GDP of $22.38 billion, crime costs represent between $112 and $217 million (0.50 to 0.97 percent of GDP), an average cost of $20.99 per drinker. Table 28: Cost of Crime Cost Percentage of GDP (%) 2014 US$ (million) Per drinker* US$ Upper Bound 0.97 217 $27.70 Average 0.74 164 $20.99 Lower Bound 0.50 112 $14.28 *Estimate using target population of drinkers 7,837,039 at the national level. Source: Authors 57 Using WHO data we apply an AAF of 0.37 for men and 0.07 for women to determine that completing SBIRT will reduce 0.0138 percentage-points in men’s (from 0.038 to 0.024 percent) and 0.0026 percentage-points (from 0.038 to 0.035 percent) in women’s probabilities of dying in a crime incident. Reductions in alcohol attributable mortality risk for crime calculations: 0.038*0.37=0.0001382 0.0138 percentage-point (men) 0.038*0.07=0.000026 0.0026 percentage-point (women) AVOIDED COSTS OF ROAD TRAFFIC ACCIDENTS Drinking and driving increases the risk of a crash and the probability of death or serious injury. This risk increases with higher levels of blood alcohol concentration (BAC). For a person with a BAC level of 0.01g/100ml the crash risk is almost five times higher than for someone with a BAC level of zero (WHO 2013). In 2010, the police recorded 1,388 road traffic fatalities across Zambia (Zambian Road Safety Trust 2014). Death tolls indicate that 46 percent of fatalities were pedestrians, 30 percent passengers, 8 percent drivers, 13 percent cyclists, and 3 percent riders of 2- or 3-wheelers motorized vehicles. However, evidence for relating alcohol and traffic accidents is limited because authorities do not enforce alcohol testing. Non-fatal crash injuries are usually not reported or documented. This results in a lack of reliable data regarding Zambia’s traffic accidents involving impaired drivers with BAC levels over the national limit of 0.08g/100ml, with very low enforcement of the national drink-driving law (WHO 2013). According to Bouchery et al. (2011), the costs of alcohol-related road traffic accidents in the United States represent 0.23 percent of GDP in 2006. These costs include damages in crash, lost of productivity, and correction costs associated with alcohol. In another study, 58 Thavorncharoensap et al. (2010) estimate that the economic cost of alcohol consumption in Thailand represented 0.023 percent of GDP in 2006. Matzopoulos et al. (2014) calculated the cost of alcohol-related traffic accidents in South Africa as 0.98 percent of GDP in 2007. According to the WHO, Zambia had 337,513 vehicles registered in 2010. South Africa has seven times the number of vehicles per 100,000 people than Zambia. Using these estimates as reference, alcohol-related traffic accidents in Zambia cost between $56 and $224 million, an average of $17.85 per drinker. Table 29: Cost of road traffic accidents Drinking State Percentage of GDP (%) 2014 US$ (million) Per drinker US$ Upper Bound 1.00 224 28.56 Average 0.63 140 17.85 Lower Bound 0.25 56 7.14 *Estimate using target population of drinkers 7,837,039 at the national level. Source: Authors We also predict that reducing alcohol consumption will decrease the probability of dying in a road traffic accident. According to the WHO, 41.7 per 100,000 men and 20.4 per 100,000 women die in road traffic accidents every year and the AAF are 0.05 for men and 0.02 for women. Based on these estimates, a successfully treated patient will reduce his or her probability of dying by 0.002 percentage-points, from 0.042 to 0.040 percent for men and by 0.0005 percentage-points from 0.020 to 0.015 percent for women. Reductions in alcohol attributable mortality risk for traffic accidents calculations: 0.00042*0.05 = 0.00002 0.002 percentage-point (men) 0.000204*0.02 = 0.000005 0.0005 percentage-point (women) 59 FAMILY COHESION AND DOMESTIC VIOLENCE For the purpose of our analysis we decided not to take into consideration domestic violence costs. First, many experts argue it is difficult putting a “cost” on the physical and psychological damage of violence against women and children. Additionally, most of the studies estimating costs of domestic violence have been conducted in industrialized countries, including costs in health care, social services, criminal justice, and businesses. For example, costs of domestic violence in the United States range between $3.5 and $12.6 billion a year. In the United Kingdom these costs added to $30.6 million (Duvvury et al. 2004). Estimating costs in developing countries is a challenging task because social norms enable domestic violence, creating a “culture of silence” among victims. Therefore, incidents are not reported and there is no utilization of services. In Zambia, 98 percent of women working in rural areas and 80 percent in urban areas are in the informal sector. Women perform most of the unpaid household labor, which makes it difficult to determine lost or reduced income attributable to domestic violence. According to a 2007 national survey, women whose husbands drink frequently are more likely to have experienced spousal violence. In this study, 77 percent of married women whose husbands get drunk frequently reported to have experienced physical, sexual, or emotional violence (CSO et al. 2009). Given the uncertainty to determine the actual number of domestic violence cases that can be attributed to alcohol consumption we did not monetize the avoided costs that SBIRT could provide if successful treating patients. We assume that omitting these costs will not substantially underestimate the net benefits of the program. 60 APPENDIX K: IMPLEMENTATION COSTS SALARY OF COMMUNITY HEALTH WORKER While the salary and benefits of the health educator or clinician implementing SBIRT typically comprises the majority of the implementation cost estimate, the proposed model in Zambia would use CHWs to implement the screenings and brief interventions, which would change the cost distribution. CHWs are compensated very differently across the country, depending on the health area and the organization for which they are working. The CHWs that would participate in the SBIRT pilot have worked with the Lutheran Health and Development Program in Mwembeshi since 2003. They have been compensated by in-kind donations from the community, bags of rice, and most recently a bicycle to facilitate home visits with an estimated value of $137 (Fretz 2011). On the national level, the Ministry of Health has proposed using CHWs to compensate for the shortage in public health workers the country faces. CHWs working through the Zambian Ministry of Health would receive roughly $295 per month (2010 dollars), for full-time, formalized work through the Ministry of Health program (Ashraf and Kindred 2011). To determine the costs associated with using CHWs to implement diagnostic screening and brief interventions, a minute wage was calculated for both the CHWs in the pilot program and the CHWs that would be used in the scaled-up program in Central Province. As previously mentioned, CHWs in the pilot are not formal workers, so there is the assumption that their work will vary through the year, based on farming season demands. A survey of CHWs in Zambia conducted by Ashraf and Kindred found that CHWs work an average of 18 hours per week and a third of the respondents reported periods of inactivity or extended leaves during their time as a CHW. To account for hours the CHWs are inactive or may be dedicating to other health needs, 61 we estimate that they would each be dedicating roughly 9.5 hours of their time per week to SBIRT related activities. The CHWs working for the province level project are assumed to be working full-time and dedicating all 40 hours to SBIRT activities. CHW minute wage calculations: Pilot: $137 yearly salary / 52 weeks / 9.5 hours / 60 minutes = $0.0037 minute wage Province: $3746 yearly salary/ 52 weeks / 40 hours / 60 minutes = $0.03 minute wage SUPPLIES The screening and brief interventions supply costs are calculated as per patient costs using estimates from Fleming et al. (2000). The supply costs are the costs incurred for one screening or one brief intervention. Dollar estimates from the study were converted to 2014 dollars and calculated as follows: Initial screening supply costs $3107/ 8962 patients = $0.35 per patient Full screening supply costs $514/ 1481 patients = $0.35 per patient Intervention supply costs $1086/ 392 patients = $2.77 per patient SCREENING AND BRIEF INTERVENTION The amount of time needed to execute SBIRT components will vary in program implementation. To account for this variation, we calculate upper and lower bounds for the screening and brief intervention costs. The standard time for the initial screen is set at 15 minutes, with an additional 7.5 to 15 minutes of preparation and transition between patients. The standard time for the comprehensive screen is 20 minutes, with an additional 10 to 20 minutes of preparation and wrap-up per patient. The following calculations take into consideration the CHW minute wage, and the time and necessary supplies per screening. 62 Initial Screening-Only Cost Calculations: Upper Bound: ($0.0037 wage) * (15 min of screening + 15 min of prep) + ($0.35 supplies) = $.461 Lower Bound: ($0.0037 wage) * (15 min of screening + 7.5 min of prep) + $0.35 supplies = $ 0.433 Comprehensive Screen Cost Calculations: Upper Bound: $0.0037 wage * (20 min of screening + 20 min of prep) + $0.35 supplies = $0.498 Lower Bound: $0.0037 wage * (20 min of screening + 10 min of prep) + $0.35 supplies = $0.461 Summation of Initial and Comprehensive Screen: Upper Bound of Both Screenings: $0.461 + $0.498= $0.959 Lower Bound of Both Screenings: $0.433+ $0.461 = $0.894 Brief Interventions Cost Calculations: Only those drinkers identified as “at risk” or “harmful” would receive brief interventions from the CHWs. One intervention lasts 15 minutes, with preparation time varying from 30 to 60 minutes. In previous SBIRT implementations, “at-risk” drinkers receive 1 to 3 interventions and “harmful drinkers” receive 3 to 6 interventions. To account for the variation of potential treatment, upper and lower bounds were calculated, again taking into consideration the CHW minute wage, and the time and necessary supplies per intervention. Costs are calculated per drinker, assuming 100 percent completion of the outlined interventions. 63 “At-Risk” Calculations: Upper Bound: (($0.0037 wage) * (15 min + 60 min prep) * 3 treatments) + ($2.77 supplies * 3 treatments) = $9.14 Lower Bound: (($0.0037 wage) * (15 min + 30 min prep)) * 1 treatment) + ($2.77 supplies * 1 treatment) = $2.94 “Harmful” Calculations: Upper Bound: (($0.0037 wage * (15 min + 60 min prep)) * 6 treatments) + ($2.77 supplies * 6 treatments) = $18.26 Lower Bound: (($0.0037 wage * (15 min + 30 min prep)) * 3 treatments) + ($2.77 supplies * 3 treatments) = $8.81 Screening and Intervention Costs by Drinking Category: Following the SBIRT programmatic flowchart, the average costs per drinker their corresponding screenings and interventions are as follows: Table 30: SBIRT Costs per Drinker Drinking State Average Range Low Risk $0.9265 $0.0325 At Risk $6.965 $3.135 Harmful $14.46 $4.76 Dependent $0.9265 $0.0325 Pilot 64 Low Risk $2.538 $0.2625 At Risk $9.59 $5.47 Harmful $21.24 $8.88 Dependent $2.538 $0.2625 Province Source: Authors INITIAL AND FOLLOW-UP TRAINING COSTS SBIRT has never been implemented in Zambia, meaning the CHWs expected to screen and treat community members would have to be trained. Training costs are calculated for the initial training, and also for a follow-up training course at year two. CHW unit cost estimates for training were provided by Ashraf and Kindred (2011). They calculated the initial and follow-up training costs for full-time, Ministry of Health CHWs. The ratio of the pilot CHW wage to the Ministry of Health CHW wage was used to calculate a per CHW cost of training for the pilot program: Province CHW minute wage / Pilot CHW minute wage = 0.03/0.0037 = 8.108 Unit cost of one day of initial training for Province level project = $218.75 Unit cost of one day of initial training for Pilot level project = $27.00 Unit cost of one day of follow-up training for Province level project = $132.90 Unit cost of one day of follow-up training for Pilot level project = $16.39 In consultations with Dr. Paltzer, we estimated that the initial training of CHW for the pilot would last 7 days and transportation costs for the LHA Trainers traveling in from Lusaka should be included. Using WIPHL reports on training full-time health educators dedicated to SBIRT, we are assuming at the province level, the initial training will last 10 days, with all 65 relevant costs included in the CHW unit cost estimate. We are estimating that an official followup training at year two would last 3 days for both the pilot and province level projects. Transportation costs: 140mi/(17mi/gallon) = 4.12 gallons*$6.09 = $25.01/round trip Table 31: Training Costs Unit training Number of cost ($ per CHW CHW, per day) Length of training (Days) Transportation Total ($) costs ($) Initial 27.00 25 7 25.01 4,750.01 Follow-up 16.39 25 25.01 1,254.26 Initial 218.70 402 10 n/a 879,735 Follow-up 132.90 402 3 n/a 160,277 Pilot Province Source: Authors RECURRENT TRAINING AND SUPPORT COSTS Costs are incurred in the provision of regular technical and social support to the community health workers. Technical and social supports have been widely acknowledged as being integral to the success of community health worker programs. In the context of the SBIRT program currently being considered, this support would include costs for the Lutheran Health and Development Program, as the primary trainer, as well as in the local clinic staff, which will be providing additional technical support. To approximate these costs, we rely on data provided in a technical report prepared by the Earth Institute at Columbia University. This report looks 66 specifically on the establishment of CHW systems in developing countries as a means to work toward achieving the millennium development goals. The aforementioned report provides a breakdown of the average yearly expenditures for community health worker programs for Zambia. Adding the annual training and management costs from this accounting and dividing them by the number of CHWs, we calculated a value for the expected support costs per health worker. For the Central Province estimate, there is an additional administrative support cost. This takes into account the administrative overhead for running a larger program, including the additional burden of managing salaries. This results in an estimated cost of $123.38 per CHW for the pilot program and $298.98 per CHW for the Central Province program. As this is based on national level government data, rather than data specific to the Lutheran Health and Development Program and clinics anticipated to provide support, we factored in a high degree of uncertainty into the Monte Carlo simulations. Specifically, we included 50 percent uncertainty on either side of the point estimate. 67 APPENDIX L: VALUE OF A STATISTICAL LIFE (VSL) In order to monetize certain costs and benefits, including those associated with health, crime, and individual welfare, analyses routinely use figures known as the value of a statistical life (VSL). A VSL is the hypothetical amount an individual would be willing to pay in order to decrease the aggregate risk equal to a single reduction in fatalities. It is not a monetary value of actual life, but rather the reduction in risk associated equal to one life (Viscusi & Aldy 2003). Studies estimating the willingness to pay for VSL have used a variety of methods. Recent federal guidance on the use of statistical life in cost-benefit analyses has encouraged the use of studies of employment decisions based on risk (Department of Transportation 2012). This research, known as “hedonic wage studies” generally use measures of income, risk of death, morbidity, and other factors associated with wage to estimate the change in salary due to increases of occupational risk. A 2012 meta-analysis of recent studies, suggested by a federal panel of cost-benefit experts, reveals a VSL of $9.23 million for workers in the United States, in 2013 USD. In addition to the variables mentioned above, estimated a VSL for developing countries must also take into account the different use of financial resources associated with lower incomes and shortened life expectancy. Recent research for the World Bank suggests that individuals in low-income countries will prefer present consumption of income in favor of future consumption, given greater uncertainty about life expectancy and quality of life (ICF International, Robinson, & Hammitt 2009). While US regulations prefer an income elasticity of 1.0 in transferring estimates of VSL, estimates greater than 1.0 are more appropriate for low-income countries given slower growth. Because VSL studies are routinely developed for high-income countries, 68 we have adopted the following methodology for transferring current VSL estimates for the United States to a Zambia-appropriate rate. Viscusi & Aldy (2003) provide the following ordinary-least-squares regression for mean VSL π = π + π½i + γr + δπ ! where V is ln(VSL), i is ln(income), r is mean risk, β γ δ are coefficients, and α is a constant. Income is measured by median household income, as estimated by the Federal Reserve of St. Louis, and the World Bank. Mean risk is estimated by danger or risk in the workplace, as measured by occupational injuries and fatalities (Hämäläinen et al. 2006). We utilized the values as in Table 32. Table 32: Values Used in VSL Calculation Value Value VSL (ln) $9,300,000 (16.05) α, constant 11.22 β, coefficient for ln(income) 1.50 γ, coefficient for mean risk -0.053 δ, coefficient for mean risk squared 0.00016 Gross national income, US (ln) $53,960 (10.896) Gross national income, Zambia (ln) $3,070 (8.029) Occupational fatality rate, US 5.20 deaths per 10,000 workers per year Occupational fatality rate, Zambia 19.80 deaths per 10,0000 workers per year Source: Authors Using the value and equations above, we calculate the difference between ln(VSL) for the United States and Zambia: United States: π(ππ) = 11.22 + 1.50 ∗ 10.896 + −0.053 ∗ 5.20 + 0.00016 ∗ 5.20! π(ππ) = 27.29 69 Zambia: π(π) = 11.22 + 1.50 ∗ 8.029 + −0.053 ∗ 19.80 + 0.00015 ∗ 19.80! π(π) = 22.30 π ππ − π π = 4.99 This represents the change in ln(VSL) transferred from the United States to Zambia. To find the figure in current dollars, we subtract 4.99 from our adopted VSL, and calculate. ln πππΏ! = 16.06 − 4.99 πππΏ! = 63081.12 We round this figure to attain $63,000 for the value of a statistical life in Zambia. Value of a statistical life year (VSLY) is calculated by dividing the VSL by an annuity factor, based on the estimated remaining years of life for the average citizen, and an appropriate discount rate (Boardman et al. 2010). Parameters used in calculating the annuity factor are described in Table 33 (WHO 2014; CSO 2013). Table 33: Values Used in VSLY Calculation Value Value VSL $63,000 Discount rate 3.5 percent (standard model) Average age, male 17 (rounded from 17.1) Average age, female 17 (rounded from 17.3) Years of life remaining, male, age 17 48 Years of life remaining, female, age 17 50 Source: Authors 1 π΄πΉ = × [1 − 1 + π π 70 ! !"!!"# ] where AF is the annuity factor, i is the discount rate, LE is life expectancy, and age is average age. At the standard 3.5 percent discount rate used in our primary model and calculations, we complete the following calculations: π΄πΉ!"#$%" = 1 × 1 − 1 + 0.035 0.035 π΄πΉ!"#$%" = 1 × 1 − 1.035 0.035 ! !"!!" ! !" π΄πΉ!"#$%" = 23.46 π΄πΉ!"#$ = 1 × 1 − 1 + 0.035 0.035 π΄πΉ!"#$ = 1 × 1 − 1.035 0.035 ! !" ! !"!!" π΄πΉ!"#$ = 23.09 Dividing the VSL by the annuity factor for each gender provides a VSLY value of $2,695 for women, and $2,798 for men within the standard model. 71 APPENDIX M: QUALITY ADJUSTED LIFE YEARS (QALY) Implementation of the SBIRT program may result in quality-of-life improvements for the population that successfully reduces its levels of drinking. The improved life quality stems from reduction in diseases and conditions related to high levels of alcohol consumption, such as heart disease, stroke, liver cirrhosis, and cancer. Several studies have attempted to calculate the improvement in quality of life due to reduction in drinking. The appropriate measurement for this change is quality adjusted life years (QALY). This measure captures the intangible benefits of improved health (Broome 1993), rating quality of health in intervals from 0 to 1. Therefore a year of perfect health and two years of health rated at .5 equal 1 QALY. It is difficult to transfer estimates across studies, as the figures are calculated for certain populations, diseases, and estimates of quality of life. For example, Solberg et al. (2008) completed a meta-analysis of randomized control trials to calculate the increased benefits from alcohol interventions in primary care settings estimated at 0.052 QALYs per person and completed screening. However, this figure does not account for referral to more intensive treatment, as called for in the SBIRT model. We have adopted figures from Mortimer & Segal, who used computer simulations to estimate QALY gains due to different alcohol treatment protocols, as given to a population of “problem drinkers” in Australia (2005). Treatments included brief interventions, a series of brief interventions, intensive counseling utilizing psychotherapeutic methods such as cue exposure and reflective listening. In the absence of a complete protocol, we assume that inpatient treatments, such as the five-day SHARPZ residential treatment in Lusaka, utilize similar psychotherapy during the course of treatment. 72 We adopted the QALY estimates from the Mortimer & Segal study for use in the Monte Carlo Simulation, as specified in Table 34. All QALY estimates were applied using a triangle distribution centered on the mean QALY gain, and ranging from 0 to the maximum figure. Table 34: Estimates of change in QALY after SBIRT treatment Treatment Population Average change in QALY (minimum – maximum) No treatment Abstainers and low drinking 0 level Brief intervention At-risk male drinkers .091 (0-.225) Brief intervention At-risk female drinkers .125 (0-.225) Series of brief interventions Harmful male drinkers .091 (0-.405) Series of brief interventions Harmful female drinkers .125 (0-.405) Series of psychotherapeutic treatments Likely dependent male and female drinkers .115 (0-.232) Source: Authors 73 APPENDIX N: HUMAN IMMUNODEFICIENCY VIRUS (HIV) Current literature on brief interventions and SBIRT do not quantify any potential improvements to health arising from reduction attainment of human immunodeficiency virus infection and acquired immune deficiency syndrome (HIV/AIDS). Given the high prevalence of HIV in Zambia (12.5. percent) and the relationship between risky alcohol consumption and HIV transmission, we include this as a separate benefit to successfully treated drinkers (UNAIDS 2014; Kalichman 2007). Benefits will accrue only to individuals who have not yet acquired HIV at the time of screening and treatment. To account of this, we utilize the incidence rate of HIV acquisition, currently .8 percent (National Council on AIDS 2014). This figure is the base probability of a new case of HIV within our screened population. Drinkers have an increased probability of acquiring HIV, for example due to decreased adherence to safe sexual practices, and less utilization of pre-exposure prophylaxis and other measures. Using data collected in Kenya, we estimate the increase probability by multiplying the incidence rate by the odds in Table 34 (L’Engle et al. 2014). Table 35: HIV Acquisition Odds by Drinking State Drinking State Increased Zambia HIV Probability Rate Incidence Abstainers 1 .008 Low-Risk 1 .008 At-Risk 2.66 .008 Harmful 9.64 .008 Likely Dependent 4.34 .008 Source: Authors Odds * Incidence .008 .008 .02128 .07712 .03472 A successfully treated individual will reduce their risk of acquiring HIV to the incidence rate of abstainers and low-risk drinkers, that is, the incidence rate of Zambia. We monetize this reduction in risk as the avoided cost of living with HIV, as estimated in QALYs. As quality of 74 life measures are sensitive to life expectancy, population characteristics, and available medical treatment, figures vary across HIV treatment types and countries where studies are performed (Farnham et al. 2013). Recent studies show that the reduced life expectancy for a male infected with HIV is around 80 percent of average age expectancy: 58 for women, 55 for men (CSO 2013). Based on this figure, we adopt Hutchinson et al. (2010) for avoiding HIV infection, which found a QALY gain 6.43 of based on computer simulations of expected years-of-life-lost of 12.6 with a discount rate of 3 percent. 75 APPENDIX O: HISTOGRAMS FIGURE 7: NPV, PROVINCE, 2% DISCOUNT SBIRT in Central Province 10,000 Trials Mean: $29,076,000 500 0 260 280 300 320 340 Net Present Value, in Hundred Thousands of US Dollars Discount rate 2%, success rate 20%. 100% of trials were positive. Source: Authors 76 Frequency 1000 FIGURE 8: NPV, VILLAGE, 2% DISCOUNT SBIRT in Mwembeshi 10,000 Trials Mean: $163,300 1000 600 400 200 140 150 160 170 180 Net Present Value, in Thousands of US Dollars Discount rate 2%, success rate 20%. 100% of trials were positive. Source: Authors 77 0 190 Frequency 800 FIGURE 9: NPV, PROVINCE, 6% DISCOUNT SBIRT in Central Province 10,000 Trials Mean: $48,621,000 500 0 400 450 500 550 600 Net Present Value, in Hundred Thousands of US Dollars Discount rate 6%, success rate 20%. 100% of trials were positive. Source: Authors 78 Frequency 1000 FIGURE 10: NPV, VILLAGE, 6% DISCOUNT SBIRT in Mwembeshi 10,000 Trials Mean: $273,600 1000 600 400 200 240 260 280 300 Net Present Value, in Thousands of US Dollars Discount rate 6%, success rate 20%. 100% of trials were positive. Source: Authors 79 0 320 Frequency 800 FIGURE 11: NPV, PROVINCE, 10% SUCCESS SBIRT in Central Province 10,000 Trials Mean: $21,112,000 500 0 180 200 220 240 260 Net Present Value, in Hundred Thousands of US Dollars Discount rate 3.5%, success rate 10%. 100% of trials were positive. Source: Authors 80 Frequency 1000 FIGURE 12: NPV, VILLAGE, 10% SBIRT in Mwembeshi 10,000 Trials Mean: $115,500 500 90 100 110 120 130 140 Net Present Value, in Thousands of US Dollars Discount rate 3.5%, success rate 10%. 100% of trials were positive. Source: Authors 81 0 Frequency 1000 APPENDIX P: PARAMETERS SUMMARY Table 36: Parameters Treated as Uncertain Source: Authors Parameter Point Estimate ($) Range Mwembeshi [Central Province] Mwembeshi [Central Province] Low risk 0.93 [2.54] 0.06 [0.53] At Risk 6.70 [9.59] 6.26 [10.94] 14.50 [21.24] 9.52 [17.76] 0.93 [2.54] 0.06 [0.53] Health Care System Cost 16.30 1.64 Uniform Clinic Support Costs $3,085 [49,600] [120,200] $3,085 [49,600] [120,200] Triangular Female 568 1,137 Uniform Male 455 911 20.99 13.42 Triangular 17.85 21.42 Triangular Abstainer 0.26 0.18 Triangular Low risk 0.61 0.53 At Risk 1.10 0.40 Harmful 1.80 1.30 Dependent 61.20 64.00 Abstainer 0.57 0.38 Low risk 1.30 1.10 At Risk 2.50 3.40 Harmful 3.90 2.80 119.30 110.5 Abstainer 0.07 0.05 Low risk 0.17 0.14 At Risk 0.31 0.29 Harmful 0.49 0.36 Dependent 26.00 35.80 Abstainer 0.24 0.15 Low risk 0.55 0.46 Screening and Brief Intervention Costs (per drinker) Harmful Dependent Administrative overhead Social Network Costs (per drinker) Avoided Costs of crime (Per drinker) Avoided Costs of road traffic accidents (Per drinker) Patient Opportunity Cost Female Formal Rural Urban Dependent Informal Rural Urban 82 Distribution ($) Triangular Triangular Male Formal Rural Urban At Risk 1.00 0.79 Harmful 1.60 1.20 Dependent 56.40 60.10 Abstainer 0.32 0.21 Low risk 0.75 -0.44 At Risk 1.40 1.24 Harmful 2.20 1.60 Dependent 72.70 73.30 Abstainer 0.81 0.56 Low risk 1.90 1.60 At Risk 3.50 3.20 Harmful 5.50 4.10 165.10 148.3 Abstainer 0.11 0.08 Low risk 0.26 0.23 At Risk 0.48 0.45 Harmful 0.75 0.53 Dependent 33.00 41.4 Abstainer 0.35 0.23 Low risk 0.46 0.69 At Risk 1.50 1.42 Harmful 2.40 1.70 76.80 76.74 Low risk 4.51 3.20 At Risk 20.94 14.83 Harmful 20.99 14.87 Dependent 60.65 42.94 Low risk 9.89 7.00 At Risk 45.93 32.52 Harmful 46.05 32.60 133.02 94.2 Low risk 1.12 0.59 At Risk 5.21 2.74 Harmful 5.22 2.74 15.08 7.94 Low risk 3.69 1.94 At Risk 17.14 9.02 Harmful 17.18 9.05 Dependent Informal Rural Urban Dependent Absenteeism Avoided Costs Female Formal Rural (per drinker) Urban Dependent Informal Rural Dependent Urban 83 Triangular Dependent Male Formal Rural Urban 49.64 26.12 Low risk 5.58 3.95 At Risk 25.90 18.34 Harmful 25.97 18.38 Dependent 75.01 53.12 Low risk 13.98 9.90 At Risk 64.90 45.96 Harmful 65.07 46.08 187.97 133.12 Low risk 1.73 0.91 At Risk 8.02 4.22 Harmful 8.04 4.23 23.22 12.22 Low risk 5.44 2.86 At Risk 25.26 13.30 Harmful 25.32 13.33 Dependent 73.15 38.51 Low risk 1.73 4.98 At Risk 8.01 23.11 Harmful 8.03 23.17 23.20 66.94 Low risk 3.78 10.92 At Risk 17.57 50.69 Harmful 17.62 50.82 Dependent 50.89 146.81 Low risk 0.00 0.00 At Risk 0.00 0.00 Harmful 0.00 0.00 Dependent 0.00 0.00 Low risk 0.00 0.00 At Risk 0.00 0.00 Harmful 0.00 0.00 Dependent 0.00 0.00 Low risk 2.13 4.98 At Risk 9.91 23.11 Harmful 9.93 23.17 28.70 66.94 5.35 10.92 Dependent Informal Rural Dependent Urban Presenteeism Avoided Costs Female Formal Rural Dependent Urban Informal Rural Urban Male Formal Rural Dependent Urban Low risk 84 Triangular Informal Rural Urban Health benefits (*measured in change in quality adjusted life years per person) Female Male At Risk 24.83 50.69 Harmful 24.89 50.82 Dependent 71.91 146.81 Low risk 0.00 0.00 At Risk 0.00 0.00 Harmful 0.00 0.00 Dependent 0.00 0.00 Low risk 0.00 0.00 At Risk 0.00 0.00 Harmful 0.00 0.00 Dependent 0.00 0.00 Abstainer 0.00 0.00 Low risk 0.00 0.00 At Risk 0.13 0.23 Harmful 0.13 0.23 Dependent 0.13 0.23 Abstainer 0.00 0.00 Low risk 0.00 0.00 At Risk 0.91 0.41 Harmful 0.91 0.41 Dependent 0.12 0.23 Table 37: Parameters Treated as Certain Source: Authors Parameter Training Costs Initial Follow-up Point Estimate Mwembeshi [Central Province] $4,750 [$879,735] Sensitivity Analysis Performed No $1,254 [$160,280] CHWs 25 [402] No Participation Rate 65% Yes HIV Incidence (Zambia) Relative Risk of HIV infection 0.8% No Abstainers 1 No Low-risk 1 Discount Rate At-risk 2.66 Harmful 9.64 Likely dependent 4.34 3.5% 85 Yes Triangular Success Rate 20% Yes VSL $63,000 No Female $2,695 No Male $2,798 VSLY Population (over age 15) Wage Female Formal Informal Male Formal Informal 3,600 [412,645] No Rural $2,405 No Urban $5,275 Rural $657 Urban $2,165 Rural $2,975 Urban $7,455 Rural $1,013 Urban $3,191 86