Recent research on economic development has identified health as one of the most significant factors separating wealthier nations from poorer ones.  Macroeconomic modeling suggests that health has a substantial impact on per capita GDP growth and is a critical input to economic development. 1 Malaria alone has been estimated to impose a 1.3 percent penalty on growth per year. 2 The projected impact of HIV/AIDS on development is equally dire; many African countries are expected to lose 10-20 percent of GDP by 2010.3 Tuberculosis (TB) will kill 35 million people in the first two decades of this century, primarily working-aged adults. 4. 5  Individually and together, these three diseases are critical to understanding the effects of health on economic development in the 21st century.
          Despite the growing recognition of the macroeconomic implications of health, the microeconomic mechanisms by which a healthier population contributes to a country's economic growth are poorly understood. 6  Among the potentially important pathways requiring further study is the impact of adult morbidity on labor productivity.  In other words, how much less productive are sick adults than healthy ones?
This study is a 2 year pilot study using data that is currently being collected in Rift Valley Province, Kenya. This preliminary analysis is part of an ongoing retrospective cohort study on the impact of common adult diseases (malaria, HIV/AIDS, tuberculosis) on labor productivity of tea plantation workers.
The main study addresses two key research questions:

PRODUCTIVITY
1)   What is the impact of malaria, HIV/AIDS, and tuberculosis on individual labor productivity?  We will answer this question by assessing the relationship between disease and the daily output of tea leaf pluckers working for two large tea companies in western Kenya. Changes in productivity resulting from both a plucker's own illness and that of a spouse (due to the need for care-giving, funeral  attendance, etc.) will be analyzed.
 
ECONOMIC DEVELOPMENT
2)   To what extent does adult morbidity hinder economic development through its impact on wages and profits?  Using the results of objective (1) above, we will evaluate the consequences of the three  diseases for workers' wages, household income, and the expenses and revenues of the two agricultural companies that are participating in  the study.
STUDY SITE
The study site is a set of 32 tea estates owned by two tea companies located in Kericho District in Rift Valley Province in the highlands of western Kenya.  The companies are large agro-industrial firms that raise and process tea, primarily for export.  One company employs approximately 11,000 tea pluckers, the other approximately 15,000.  Each company maintains a central hospital and a system of dispensaries and clinics.  The tea companies provide full medical care to all workers and dependents free of charge at the company dispensaries, clinics, and hospitals and provide paid sick leave.
STUDY POPULATION
Cases are pluckers who have died at company health facilities of AIDS-related conditions, tuberculosis, malaria or other natural cause deaths, pluckers with a spouse who died at company health facilities of any one of these conditions, and pluckers who were medically retired due to one of these conditions. Controls are pluckers who were working in the same fields at the same time as the cases.
DEATHS:  Primary data collection was initiated with the development of a database of all natural cause deaths  from both of the company hospitals.
EXCLUSIONS:  From this initial database children under 18 and all workers who were not directly involved in plucking (and therefore not paid by the kilogram plucked) are removed from the data set.
CONTROLS:  Using the set of known cases, suitable controls are chosen from the historical work gang records at the tea sites. Controls are chosen from other pluckers from the same work gangs (units) as the case pluckers.  For each illness index case, 4 controls matched by work gang and day will be selected.
TIME FRAME: Data is extracted for the four years preceding death or medical retirement.  For the death of a spouse, data is also extracted for the two months following the death, when productivity is likely to be affected by funeral attendance and other family needs.
METHOD OF DATA EXTRACTION: Data is collected directly from the handwritten registers of daily productivity. A team of data entry staff enter the data into a hand held data collection device comprised of a Palm Pilot with spreadsheet software and a keypad attachment. Data can then be imported onto a computer and cleaned.
The Impact of Morbidity on Labor Productivity
in Rift Valley Province, Kenya
 
M P Fox , S B Rosen, W B MacLeod, M Bii, L H Elson, K M Wasunna, J L Simon
Kericho is approximately 300 kilometers (4 hours driving) from Nairobi.
Data Analysis
n.a.
10
Medically retired prior to death
n.a.
4
Spouses as cases
7.2 (2.6)
5.6 (3.2)
Mean years of service
40 (8.8)
39.5 (8.7)
Mean age
71.9%
72.5%
Percent male
Control (n=175)
Case (n=43)
 
Population Demographics                   [Freq / Mean (std dev)]
 
Differences in Plucking and Daily Wage
Mean Differences in Kilograms Plucked
Mean Differences in Daily Wage
Other Findings
Differences in Kilograms (kg) Plucked
After adjusting for other covariates, we found a trend towards a difference but it did not reach conventional statistical significance.  Controls plucked 36.9 kg on average while cases (deaths) plucked 33 kg (p=.056) over the four years. At the standard union wage rate rate of 4 shilling per kilograms, this equates to a difference of 148 shillings as an average daily wage among controls, and an average daily wage among cases of 132 shilling over the four years of observations.
In the last year of observation before a death, the difference was even greater and became statistically significant. Controls had an average of 36.5 kilos plucked, while cases only averaged 29.7 kilos per day (p=.025). This translates to an average daily wage for controls of 146 shilling, while cases earned 119 shilling.
We modeled our data using a generalized linear mixed model with daily output of kilograms plucked as our dependent variable, and estimated a regression equation of the form:
 
The b coefficients indicate the impact of each independent variable on daily output.  The individual-specific intercept (b1i) is intended to capture the effects of unobserved personal characteristics that affect productivity.  
Within our model we controlled for the following effects: experience (years of service), age, gender, which tea estate the tea plucker worked on and disease progression. We also included in our model dummy variables to indicate case/control status and a dummy variable for the matched set. We modeled time as a random effect and allowed for a random intercept in order to account for the high correlation between the repeated observations on a single plucker over the four years of observation.

We repeated our model again using only data from the last year of observation to determine the effects of disease in the final year before death. We then analyzed the data from only the cases to determine the effects of disease progression over time. Finally we looked for statistical differences in leave used and amount of absenteeism.
Methods of Data Collection
The Issue
Goals of the Study
Study Site and Population
Conclusions
.48
-1.09
Female
.0561
-3.84
Case (death)
.056
-.18
Age
.19
.39
Years of Service
P-value
Measure of Effect
Covariate
Quantity of Tea Plucked Over 4 Years Before Death (in kilograms)
.44
-1.9
Female
.025
-6.8
Case (death)
.02
-.34
Age
.10
.78
Years of Service
P-value
Measure of Effect
Covariate
Quantity of Tea Plucked in the  Year Before Death     (in kilograms)
In an unadjusted comparison of cases and controls, we found a decrease in average daily output over time among cases in the four years before death. This effect is most pronounced in the final year before death when cases plucked on average 29.4 kilos while controls plucked 35.6.
On average, cases used 17.9 more sick days (p<.0001) in the second year before death and 12.0 days (p<.0001) 1 year before death than did controls. They also used 8.5 more days of annual leave in the second year before death (p<.002) and 9.6 more annual leave days 1 year before death (p<.0009) than did controls. Cases were absent without pay 14.3 days more than controls in the second year before death (p<.0001) and 15.8 (p<.0001) more days 1 year before death.
In looking at just our cases, we found that average daily plucking decreased in the three years preceding death, going from 36.1 kilograms three years before death, to 35.4 kilograms in the second year before death to 34.6 kilograms in the last year before death.
While the current results are preliminary, they provide some of the first empirical estimates of the impact of common adult illness on labor productivity.  Workers on these tea estates in Western Kenya are experiencing a reduction in the average number of kilograms of tea plucked and an increase in the amount of absenteeism and leave days taken in the years prior to death as compared to the other members of their own work gang.  As workers are being allowed to bring helpers with them to pluck when the crop is high, and sick workers may be more likely to use helpers, the true differences between sick workers and healthy workers may be underestimated. As workers are expected to pluck a minimum of 32 kg per day, this decrease in productivity to under 30 kg in the year before death may put sick workers in jeopardy of losing their jobs. The increase in absenteeism and leave is a financial burden to the company as total productivity declines.  In addition since the companies provides medical care and hospitalization to their employees, the increasing burden of disease also affects company profitability. By measuring the financial impacts that diseases have on labor productivity, we hope to make an economic argument for greater investments in prevention and care initiatives.  If business is to really join the War on AIDS, we need to be able to make economic-based arguments for their heightened participation or they will simply shift the burden onto the public sector and households.
1  Bloom DE, Canning D: The health and wealth of nations.  Science 2000;287:1207-1209.
2   Gallup, JL and Sachs, JD. The Economic Burden of Malaria. 1998.  Center for International Development at Harvard University.
3  National Intelligence Council. The global infectious disease threat and its implications for the United States. 2000. National Intelligence Estimate 99-17D. 
4  Murray C, Styblo K, Rouillon A: Tuberculosis, in Jamison DT, Mosley WH, Measham AR, Bobadilla JL (eds): Disease control priorities in developing countries. Oxford, Oxford University Press for the World Bank; 1993:233-259.
5  WHO. Global tuberculosis control:  WHO report 2000.  WHO/CDS/TB/2000.275.
6  Collier P, Gunning JW: Explaining African Economic Performance.  Journal of Economic Literature 1999; 37:64-111.
This project was funded by a grant from the Fogarty International Center of the National Institutes of Health, Grant Number TW05605-02