State Comparison

            In order to determine the districts of focus for this study, I compared the pattern of the API of all districts within a state to that of other states for the years 1975-1984 and 1986-1995, for fourteen states.  I chose these fourteen states after eliminating all island and very small states and Haryana because the district malaria data for this state were missing a lot of information.  I did not use Gujarat and Rajasthan, although they do have fairly high API levels, because other scientists are currently investigating the relationship between climate and malaria in these states.  The fourteen states that I did compare are the fourteen with the highest average API levels after the others are eliminated.  All states below the cutoff level had an average API level of below 1.33 for the years 1961-1995.  I plotted the district API levels over time for each state and qualitatively compared them.

            Because this study intended to identify what predicts the changes in malaria rates rather than just the level of malaria, I was more interested in the variability and trends of API rather than in the magnitudes.  When doing trend graphs for all districts within each state, I found that there were a few main patterns of district API trends by state. 

           One pattern, which occurred in Himachal Pradesh, Punjab, Nagaland, Jammu & Kashmir, Assam, and Uttar Pradesh, involved high API levels in the 1970's that decreased considerably by the mid 1980's mostly to magnitudes below 20 API.  Three states, Meghalaya, Mizoram, and Tripura, only had a few districts with low API levels throughout, making the identification of any pattern in district API rates difficult.  Arunachal Pradesh had the highest API levels of the fourteen states that I analyzed.  However, because most of the high malaria prevalence occurred in one district, it did not seem to be a very interesting state on which to focus my study.

            In Orissa, the magnitude of API remained constant over time from 1975-1995, with similar up and down movements in all districts year to year.  There was a dip in API in Orissa in 1977, but all of the districts decreased by the same amount.  This dip in 1977 in Orissa, however, was not maintained and API levels rose back up in 1978 to similar levels that they had at in 1976.  1991 also showed a peak but then in 1992, the API levels resumed to what they were before.  Because of the deviation from the trend, Orissa interested me.  I assumed that if the malaria prevalence rates stayed relatively the same over through the entire period of my study, that either malaria control efforts were not implemented, or that malaria control efforts did not succeed in controlling malaria.  Malaria and climate might have been more strongly correlated here than in other states because of this apparent lack of an effect from malaria control techniques.  Orissa, however, is one of the poorest states and therefore does not have well-administered malaria control program or data collection.  This means that the malaria control efforts were most likely not fully implemented and also that the data are probably not as reliable.  Given this information and also the information that districts in Orissa are very ecologically diverse, possibly more so than Karnataka (Guha, 2000, personal communication), I decided not to focus on Orissa in this study.

            Three districts followed more explicitly the history of malaria control programs in India over this time period:  High API levels in the 1970’s with a peak in 1976 before the implementation of the MPO in 1977.  Starting in 1978, API decreased to a low in the mid-1980’s and then increased again in the 1990’s.  Maharashtra, and Madhya Pradesh, showed a sharp drop in API in 1977 after peaks in 1976.  The API trends in these states also exhibited an increase in API in a few districts in the early 1990’s, with a peak in 1991. 

Karnataka also followed the malaria control history trend in some of its districts, but there was a lot of heterogeneity and cross-over of district API trends.  A decipherable peak in 1976 was followed by a gentle decline in API levels in all districts to a low in 1984.  There were increases in the 1990s, particularly in two districts starting in 1992, without the characteristic peak in 1991.  The heterogeneity of climatic regions and the heterogeneity in API trends and climatic regions by district in Karnataka lead me to decide to focus my study on this state.  Karnataka has also been fairly politically stable and thus is likely to have more accurate data over time (Bradley, 2000, personal communication).  Finally, there were few changes in the names of the districts in Karnataka between 1975 and 1995, making the analysis over time much easier.