# Partners in Health - December 2007 review

We have published a more recent review of this organization. See our most recent review of Partners in Health.

## What do they do?

We first overview PIH's activities worldwide, showing that its "Haiti model" - a comprehensive rural health care program including a hospital, local health centers, and community health workers - is its primary program. We then focus on the application of this model in Rwanda.

### Overview of activities

Attachment A-1, Pg 1 gives a complete list of programs, with brief overviews; PIH's website has longer descriptions. Attachment B-11 gives the total expenses for each over the last few years.

The "Haiti model," described below, refers to a comprehensive program clinic. PIH operates this model in:

• Haiti, in the rural Central Plateau since 1985.
• Rwanda, at six rural sites since 2005.
• Lesotho, in the mountain village of Nohana.
• Malawi, in the village of Neno since January 2007.

Other programs include:

• A large health care program in Peru, operating since 1994.
• A program in Russia that appears to focus on research and training local technicians to respond to a tuberculosis epidemic, operating since 1998.
• A Boston program that focuses on health care for HIV patients.
• "Supported projects" (i.e., providing relatively limited support to other organizations) in Mexico and Guatemala.

The table below, taken from Attachment D-2, shows PIH's budget allocation across these programs for 2006 (actual) and 2008 (projected).

Region 2006 Expenses (actual) 2008 Expenses (proj) % of total 2006 expenses % of total 2008 expenses
Haiti $14,101,000$18,873,000 45% 36%
Rwanda $4,862,000$9,981,000 16% 19%
Lesotho $226,000$3,851,000 1% 7%
Malawi $23,000$3,428,000 0% 7%
Peru $3,786,000$4,843,000 12% 9%
Russia $4,284,000$3,113,000 14% 6%
Boston $799,000$922,000 3% 2%
Other $1,516,000$4,109,000 5% 8%

Note that the Africa programs, built on the "Haiti model," are slated to grow significantly; furthermore, "Haiti model" programs accounted for a total of 60% of PIH's budget in 2006, and are slated to account for closer to 70% in 2008. We therefore feel that by focusing on understanding the "Haiti model" in depth - as we do below - we can largely understand the approach of PIH as a whole.

### Rwanda program

PIH provided us with a draft of a report, commissioned by the Clinton Foundation, that gives an extremely thorough description of the "Haiti model" as implemented in Rwanda. Unfortunately, PIH does not have permission to make this report public, and so we do not either. We have done our best to summarize the most important information from this report, and PIH has informed us that a similarly thorough report will become publicly available at some point in the future. All references to Attachment B-6, below, are references to this report.

The basic approach of the model is to fund and staff a comprehensive rural health care program: a central hospital and several satellite health centers, with the staff and facilities necessary to handle the area's "most serious needs” including HIV/AIDS, tuberculosis, and malaria (see our problems and solutions overview for more on these diseases and their treatments). The program not only provides centralized health care, but also has a significant home-care component through community health workers (see "Accompagnateurs" below), and does some work to address other needs such as housing and credit (see "Program on Social and Economic Rights" below).

The program is eventually intended to become locally run and funded by the Rwandan government (Attachment B-9), though we do not focus on this aspect of it for reasons given below.

The following sub-sections give more detail on what is involved in starting up and staffing the program, as well as the specific medical treatments that PIH focuses on. Unless specified otherwise, all facts below come from Attachment B-6, the draft report that is not yet publicly available.

#### Startup

In the startup process, PIH researches and analyzes the area's disease burden. It is not clear to us exactly what this research consists of or how PIH chooses new areas to start programs in, but in the case of Rwanda, it appears that the program reaches a population in significant need. Attachment B-6 cites a UNICEF survey (exact reference not given) implying that the district's population has extremely poor housing1 and health.2 Startup consists of training personnel and upgrading facilities, not in any necessarily strict order: PIH states, "It is possible to commence with health centers and a community program, provided the unit is situated close enough to an adequate third party-run district hospital able to receive transferred patients." In this case, the Rwanda program started with an existing public district hospital, and performed major renovations including water, electricity and communications systems. PIH also supports several Health Centers to help reach more people. Each Health Center is equipped for HIV/AIDS and tuberculosis diagnosis and treatment, as well as maternal care.

#### Staff

PIH provides some expatriate staff in key roles, but focuses as much as possible on training locals (and having them train others). It explicitly "uses doctors sparingly and invests heavily in training nurses to be able to handle most procedures," including adult antiretroviral treatment; the sites discussed employ 53 nurses and 4 doctors.

A key part of the PIH model is its use of community health workers (accompagnateurs). Accompagnateurs perform regular home visits to patients and spend an average of 30 minutes a day with each patient. Unlike medically trained personnel such as nurses and doctors, accompagnateurs need only basic literacy and reasonable health, and can be recruited from the ranks of the unemployed; their primary role is to ensure that patients comply with their complex medical regimens, specifically the treatments for tuberculosis and HIV/AIDS (see our problems and solutions overview for more information on tuberculosis and HIV/AIDS treatments). Survey data indicates that patients are generally satisfied with this element of the program.

Attachments B-1 through B-4 were sent to us to make the case that the accompagnateur program increases compliance with complex treatments (particularly DOTS and ART); Attachment B-6 also has an extended defense of this program aspect. PIH apparently considers this a distinctive element of its organization. We can't speak to the approach's distinctiveness, since our understanding of other organizations' health care is not nearly as detailed as our understanding of PIH's, but the program is reported as taking up only around 5% of total costs, and intuitively, it seems to us like a wise investment: using relatively low-cost labor to improve compliance with medical treatments that can be quite complex.

#### Activities

PIH addresses many of the major causes of death and debilitation in Africa, generally through treatment rather than prevention. Details follow.

Attachment B-6 gives a full breakdown of PIH's hospitalizations and consultations by symptom. We reproduce the major ones:

Health issue Proportion of hospitalizations Proportion of health center consultations
Malaria 62% 47%
Deliveries and Maternal Health 12% 9%
Respiratory Tract Infections 4% 15%
Trauma 3% 2%
Other 3% 11%
AIDS 3% 1%
Diarrhea 2% 3%
Intestinal parasites N/A 8%

The dominant one (in terms of number of cases - not necessarily in terms of expenses or impact) is malaria, which is extremely common in the region (Attachment B-6 cites a Ministry of Health estimate - unsourced - that "every person will experience at least two episodes of malaria per year"). The severity of this disease varies widely and treatments range from simple drug therapy to hospitalization; we do not currently have a good understanding of the effect of malaria treatment on life outcomes.

PIH's major expenses - as well as the factors it points to most heavily as setting its health care apart - are more related to HIV/AIDS and tuberculosis (Attachment B-6). Those with serious cases are put on treatment regimens including the appropriate drugs, regular (at least daily) visits from accompagnateurs (more on these under "Staff" immediately above), and regular food packages (Attachment B-6 has more detail on food packages). Maternal care also includes drug regimens and feeding programs designed to prevent the transmission of HIV to children.

Attachment B-6 claims unusually high adherence to ART programs, observing that only 4 patients (out of over 1000) had stopped adhering, compared to a national average of 5% for ART in general (although we couldn't find the source cited). Attachment B-6 observes that out of 474 patients treated for tuberculosis, none have stopped adhering.

Finally, PIH's program includes non-treatment activities designed to "address the underlying social and economic conditions that put poor people at disproportionate risk for poor health and constitute barriers to effective treatment." These activities, which constitute a noticeable (8%) proportion of the overall budget, appear relatively simple and straightforward. The major ones are:

• Local construction and refurbishment of houses.
• Paying school fees for children "designated to be in great need, as determined by the social workers."
• Payment of government health care fees for patients.
• A small microlending program that is expected to be self-sustaining, and thus not to be a factor in the budget.

#### Sustainability

PIH stresses that its program is particularly attractive because it is expected to be eventually supported by the Rwandan government, and thus to be sustainable. Attachment B-6 gives an analysis of the feasibility of this plan, stating that supporting similar health care centers across rural Rwanda would require Rwanda's health care spending to go from around 3% of GDP to around 8%. As the report observes (using data from the WHO World Health Report 2006), this would be higher than that of almost any other country in Sub-Saharan Africa, and more comparable to a typical developed country.

Given (a) the relatively low initial buy-in of the government (around 10% of the $2.5 million startup); (b) the lack of information about similar past efforts to encourage such a large scale-up in health care spending and; (c) our wish to recommend charities based on how reliable and cost-effective their activities are, rather than based on how likely they are to pass the expenses of these activities to others, we do not factor sustainability of the program heavily into our assessment. ## Does it work? Overall, we are reasonably confident in PIH's approach, because it is: • Straightforward. Education, promotion and distribution campaigns do not clearly answer the large question of whether they're impacting people's behavior. A hospital focusing on direct treatment does not face this same problem. • Well monitored. PIH provides relatively comprehensive data on treatments given (Attachment B-6 and Attachment B-7) and incomplete data on life outcomes (particularly HIV/AIDS). • Systematic. As stated above, we feel we are able to understand a large amount of what PIH does by examining one region in depth, because many of PIH's regions explicitly aim to replicate the same model, whose implementation and strategy are clearly documented. ### Improvement in care Improvement in care addresses the question of how much better off PIH's beneficiaries are compared to how they would be with their existing system of health care. In Rwanda's case, Attachment B-6 states that Rwandan rural areas commonly have "standard" health centers (supported by the Ministry of Health) with maternity wards and some diagnostic ability; it summarizes PIH's value-added as follows: • Attachment B-6 states that the PIH program is distinguished by more and better-trained staff, superior hygiene, and superior ability to treat HIV/AIDS and tuberculosis - specifically, in the use of accompagnateurs and food packages (see above). We find this a plausible claim, in light of the poverty of the region. • Attachment B-6 states that the hospital and health centers attract significant numbers of patients from outside their districts: ~15% of the hospital's patients come from outside its district, and 10-30% of those served by the health centers come from outside their districts. This may be confirmation of PIH's superior care; PIH attributes the relatively large number of visits from outsiders to its "reputation for high quality care and the health centers' ability, unique in the immediate vicinity, to treat HIV/AIDS and TB patients." • The most quantified information we have, on this topic, is the comparison in Attachment B-6, stating that 5% of ART patients nationwide are lost to followup (though we don't know what the cited source, "TRAC report," refers to) as compared to around 0.5% of PIH's ART patients. Attachment B-8, Slides 10-24, tells a more vivid and extreme story about PIH's health care vs. the previously existing system. From these pictures, it appears as though the previously existing hospital was not just mediocre, but neglected. If the pictures are representative (and we would guess that they are) of other hospitals, PIH can safely be assumed to be bringing in a drastic improvement in the quality of health care. PIH's statement that it specifically seeks out areas with extreme need (Attachment B-9) implies that its other programs are coming into similar situations. Overall, we feel highly confident that PIH is improving the quality of local health care and therefore saving lives. ## What do you get for your dollar? We have relatively little information on life outcomes for those treated by PIH, aside from some limited statistics on HIV/AIDS and tuberculosis patients and a summary of a study on its bednet distribution program (Attachment B-10, which states that the program reduced malaria-based hospitalizations by 50%). Attempting to take a broad view of PIH's possible impact, we very roughly estimate that it can be expected to spend ~$3500 per life saved, not including non-mortality-related improvements in health.

#### Costs

The startup (2005-2006) costs for the care provided in the Southern Kayonza district were $2.5M, and PIH expects both the annual costs and the patients treated to rise significantly (resulting, overall, in falling costs-per-patient). PIH partnered with other organizations including the Ministry of Health for startup, but paid 70% of the costs itself; it anticipates continuing to play a major role in funding for the near-term, but hopes eventually to hand off the program to the Rwandan Ministry of Health (see above). When it is fully operational, PIH expects the Southern Kayonza district portion of its Rwandan program to cost around$4m per year in the long run (though there is an open question as to who will be paying for it).

#### Estimating lives saved

We cannot give a good estimate of lives saved based on PIH's specific activities, because our understanding of them varies too widely; in some cases we have a good sense for the connection between its treatments and lives saved, but in other important cases we do not (for example, PIH treats a large number of malaria cases and respiratory infections, but we do not have information on how many of these cases are life-threatening or on the outcomes for patients treated).

However, we try to get some sense of how much PIH can hope to accomplish, by comparing overall mortality rates (from relevant conditions) across different regions of the world. The idea is that by providing high-quality, comprehensive health care, PIH can hope to make mortality rates in the area it works in (poor rural area in Rwanda) move closer to mortality rates in other parts of the world with superior health care.

We recognize that this method is rough, for there are many factors that will affect mortality rates other than the existence of good healthcare: climate, living conditions, sexual behavior, prevalence of sexually transmitted diseases, transportation infrastructure (which enables people who need care to reach healthcare facilities), etc. Nevertheless, we use what follows in order to get a range of estimates for PIH's impact.

We focus on the most relevant causes of death. As our problems and solutions overview explains, malaria, diarrhea, and respiratory problems (as well as neonatal issues) primarily affect under-5 mortality; these, along with HIV/AIDS, tuberculosis, and maternal mortality, are the main causes of death that are significantly greater in Africa than in other parts of the world (see our life expectancy analysis for more on this). As Attachment B-6 shows, these conditions also account for the bulk of PIH's activities.

The table below gives both the under-5 mortality rate and the "Big three" mortality rate (annual rate of death from HIV/AIDS, tuberculosis, and maternal mortality) for a variety of regions: both rich and poor Rwanda, different parts of Sub-Saharan Africa, and finally less disadvantaged parts of the world. It is sourced as follows:

• The 2006-2007 Human Development Report for Rwanda gives the under-5 mortality rate for both the "poorest 20%" and "richest 20%" of Rwanda (note that this rate is equal to the percentage of all children who die before the age of 5; we mathematically adjust it to an annual mortality rate).
• World Health Organization data gives deaths by condition and country and the under-five mortality rate by country; we use this data to give rates for Rwanda as a whole, for the "average," "75th percentile" (i.e., lower relevant mortality than 75% of sub-Saharan African countries), and "90th percentile" parts of Sub-Saharan Africa countries, as well as other continents.
• We don't have "big three" mortality rates for "rich" and "poor" Rwanda, but we estimate them. "Rich" Rwanda is similar to the average Sub-Saharan African country in terms of child mortality, so we use the average Sub-Saharan African "big three" mortality rate as well. Our estimate of this particular number for "poor" Rwanda is just a guess, but our calculation is not very sensitive to it.
Region Under-5 annual mortality rate "Big three" annual mortality rate
Rwanda: poorest 20% 4.497% 0.500%
Rwanda: nationwide 3.766% 0.400%
Rwanda: richest 20% 2.906% 0.300%
Sub-Saharan Africa: median country 2.806% 0.300%
SSA: 75th percentile country 1.924% 0.100%
SSA: 90th percentile country 1.362% 0.000%
Middle East / North Africa 1.175% 0.038%
Southeast Asia 1.009% 0.039%
North America 0.133% 0.003%

PIH expects to serve a population of 265,000 people (Attachment B-6) in the Southern Kayonza district (the area for which PIH expects to spend $4 million/year when its program is fully operational); note that this is not the number of people treated, but rather the estimate of the entire population in the area reached by the program. We roughly estimate that ~21% of this population is under the age of five, based on the statistics for Sub-Saharan Africa provided by the Disease Control Priorities Project. By multiplying the number of people served by the mortality rates above, we estimate how many relevant deaths we would expect per year in the areas served by PIH - if conditions were similar to those of the regions above. We assume that without PIH's help, the region's conditions would be most similar to the first row in the table (the poorest 20% of Rwanda). The difference between deaths under North America-like conditions and deaths under poor-Rwanda-like conditions equals the "lives saved" by PIH, if it were able to create conditions similar to North America in the region. Finally, the "Cost per life saved" column divides the "lives saved" into$4.5 million, PIH's annual costs (see the section immediately above).

Region Under-5 annual deaths "Big three" annual deaths Total annual deaths "Lives saved" Cost per "life saved"
Rwanda: poorest 20% 2479 1049 3528 - -
Rwanda: nationwide 2076 839 2915 613 $7340 Rwanda: richest 20% 1602 630 2232 1297$3470
Sub-Saharan Africa: median country 1547 630 2177 1352 $3329 SSA: 75th percentile country 1061 210 1271 2258$1993
SSA: 90th percentile country 751 0 751 2777 $1620 Middle East / North Africa 648 79 727 2801$1606
Southeast Asia 556 82 638 2891 $1557 North America 73 6 80 3449$1305

We would guess that a country's mortality rates are driven by much more than the quality of local health care: climate, people's behavior, and existing prevalence of infectious diseases are all major factors in the causes of death we examine here. Therefore, our best guess as to what PIH can reasonably be expected to accomplish comes from row 3: if PIH can bring the quality of care where it works to the level of care in higher-income parts of Rwanda, it saves a life for about every $3500. An alternative perspective, that ends up in the same range of estimates, is to focus on "Big three" mortality, since HIV/AIDS and tuberculosis are PIH's biggest areas of focus (see above sections). If PIH can bring mortality only for HIV/AIDS, tuberculosis, and maternal care to the level of North America - but has no impact on under-5 mortality - this implies 1043 lives saved, for a cost of$4,314 per life. This estimate likely overstates PIH's impact on "big three" mortality while understating its impact on infant mortality.

We do not believe it would be realistic to make Rwanda resemble North America (or any region outside Africa) in terms of overall mortality rates, but we provide those numbers for reference, as what we believe represent an "upper bound" on PIH's cost-effectiveness.

Note that these estimates are only for lives saved; they do not include the many other improvements in quality of life (aside from lower mortality) that can come from having improved health care. Therefore, a direct comparison between this figure and Interplast's "cost per surgery" figure will unfairly favor Interplast; comparisons to PSI's bednet or condom distribution program are less clear, since PSI's programs also have strong additional quality-of-life benefits (reduced non-fatal malaria cases, reduction in unwanted pregnancies, etc.)

### The organization

Size and scope. At around $30 million in 2006 expenses, PIH is smaller than many of our other finalists, though it appears to be expanding rapidly (see below). As we explain above, much of the organization follows a fairly unified strategy of expanding its comprehensive health care model to disadvantaged parts of the world. Personnel. 4 of PIH's 11 Board members (Attachment C-1) are M.D.'s; 4 others have corporate affiliations; one has no affiliation listed; one is the Executive Director of the Equal Justice Initiative of Alabama; and the final Board member is the Executive Director, Ophelia Dahl, a founding trustee. Other major staff (Attachment C-7) include Chief Operating Officer Paul Zintl (a former Managing Director at J.P. Morgan), VP of Program Management Theodore B. Constan (former chief of staff to the President of the Massachusetts Senate), and Founding Director Paul Farmer, a Harvard Medical School professor who was profiled in Mountains beyond Mountains. Financials. The following table is constructed as follows: 2002-2005 expenses are from Attachment B-11, 2006 and 2008 revenues and expenses are from Attachment D-2, and 2004-2005 revenues are from the IRS Form 990s available on GuideStar (2003-2005).  Year Revenue Expenses 2002 ?$12,025,174 2003 ? $14,546,690 2004$12,871,870 $14,678,167 2005$36,202,484 $22,820,668 2006$32,896,357 $31,104,000 2008 (proj.)$47,736,000 $51,709,000 PIH appears to have been expanding rapidly, and to anticipate significant growth in the future as well. As of the most recent available audit (end of 2006), PIH held about$30 million in assets, or about one year's worth, which is in the range we consider reasonable (neither excessively hoarding money nor in a precarious position).

As the following table (from Attachment D-2) shows, PIH's revenue sources are fairly balanced across individuals, large foundations, and government/multilateral donors (note that the "Other" line comes mostly from investment income, "gifts to capital," and "gifts reimbursing prior year's expenses"; we aren't sure what sort of giver the latter two items come from).

2006 2008
Contributions from Individuals and Family Foundations 37.9% 47.6%
Grants from Operating Foundations and Corporations 26.3% 28.0%
Grants from Government and Multilateral Donors 23.2% 23.2%
Other 12.6% 1.2%

### Conclusion

PIH is one of three finalists (the others being PSI and Interplast) whose model is relatively scalable and repeatable, and which we can therefore form confidence in from a bird's-eye view of the organization. Choosing between the three is a difficult judgment call. As the above analysis shows, PIH ranks below the strongest projects we've seen in terms of estimated "cost-effectiveness," i.e., cost to save a life.

However, the comparison is not this simple. PIH's model is extremely appealing to us because of its straightforwardness and comprehensiveness. PIH has the opportunity to integrate with a community and adapt to its needs in a way that Interplast (and, to a lesser extent, PSI) cannot. In addition, PIH's impact on quality of life likely goes far beyond the number of deaths averted; non-fatal health problems are still problems, and PIH's "cost per life saved" number likely understates its impact relative to that of an organization like Interplast, which treats a very discrete number of patients.

Ultimately, we rank PIH below PSI, which we believe is several times more cost-effective in terms of lives saved, and not clearly less cost-effective in terms of other benefits (since PSI's activities, like PIH's, create significant benefits aside from saving lives). We narrowly prefer PIH to Interplast, and solidly prefer it to our other finalists.

### We'd like to know more about:

• Malawi activities. Is the Malawi site the same "Haiti" model as that followed in Rwanda and Lesotho?
• Monitoring and evaluation of Haiti site. We'd expect Haiti, PIH's oldest program, to be at least as well documented as Rwanda; we'd like to see this documentation.
• Outcomes of treatment. How many of those treated have life-threatening symptoms, and how many survive after treatment?
• Likely impact of improvement in care. What evidence is available regarding the adherence to HIV/AIDS and tuberculosis treatment (and the availability of this treatment) for people with "standard" (low-income, rural) access to health care? What is the likely quality of malaria treatment and maternal care without PIH's help?

## Attachments

### A. Application and response

• Attachment C-1: Board Of Directors (currently in hard copy only)
• Attachment C-2: Letter Of Determination (currently in hard copy only)
• Attachment C-3: Mission Statement (currently in hard copy only)
• Attachment C-4: Number Of Staff (currently in hard copy only)
• Attachment C-5: Relationships with other organizations
• Attachment C-6: Sources Of Income (currently in hard copy only)
• Attachment C-7: Staff Biographies (Pg 1)
• Attachment C-7: Staff Biographies (Pg 2)

### D. Financials

• Attachment D-1: Budget 2005-2006 (currently in hard copy only)
• Attachment D-2: Budget 2006-2008

We made these notes during a February 2009 review of PIH's materials. Our most recent take is available at our 2008-2009 review of PIH.

### Update on PIH's transparency

In 2007-2008, we recommended PIH for its straightforward, well-documented approach rather than for its transparency. Most of our information came from a document that has not been cleared by the funder for public release. We are still not able to locate this document in the public domain. We would like to see more detailed information about PIH's activities and outcomes.

### Update on PIH's activities

Based on page 32 of PIH's 2008 Annual Report (PDF), we believe that PIH's health programs in Haiti and Africa remain its largest and most rapidly expanding programs. Programs in Africa, in particular, appear to have grown in their share of PIH's overall expenses. As these are the programs we focus on in our review, we feel that our review is still broadly representative of PIH's activities as a whole.

### Update on PIH's effectiveness and cost-effectiveness

PIH still does not appear to publish information on how it monitors the quality of its care and its impact on patients. We urge caution in interpreting our cost-effectiveness estimate, as it involves many assumptions, but we do not wish to revise this estimate numerically.

• 1.

"Half have leaking roofs and a third holes in the walls" and "43% of households spend over one hour every day collecting water and 14% spend two to three hours". Attachment B-6.

• 2.

"40% of children under-five years had a fever in the two weeks prior to the survey". Attachment B-6