Among patients hospitalized with AFI at the All India Institute of Medical Sciences during July 2017–September 2018, the major circulating Plasmodium species was P. vivax and malaria/DENV co-infections predominated. A high number of severe malaria cases reported to the institute are from northern India. Among the 5 Plasmodium species known to infect humans, in our study population we detected P. falciparum, P. vivax, and P. knowlesi but found no evidence of P. malariae or P. ovale. Most AFI patients in this study originated from northern India across the states of Rajasthan, Haryana, Punjab, Delhi, Uttar Pradesh, Bihar, and West Bengal. The burden of co-infecting pathogens in patients with malaria was revealed by a combination of complete blood work (peripheral blood smear analysis, rapid diagnostic testing, serum renal and liver function testing) and in-depth molecular assays (PCR amplification of Plasmodium species–specific genes followed by Sanger sequencing). We found a very high percentage of Plasmodium/DENV co-infections in our study population. This finding can be partly attributed to the highly sensitive PCR diagnostic methods used.

A recent meta-analysis of the prevalence of DENV/ Plasmodium/CHIKV co-infections spanning 7 geographic regions (southern Asia, Africa, Southeast Asia, South America, North America, the Caribbean, and the Middle East) showed that DENV/Plasmodium co-infections have been reported in 19 countries, including India; DENV/CHIKV co-infections have been reported in 24 countries including India; CHIKV/Plasmodium co-infections have been reported in 6 countries with only a single co-infection reported from India; and DENV/CHIKV/Plasmodium co-infections have been reported in 3 countries (5). According to that meta-analysis, the average reported prevalence of DENV/Plasmodium co-infection in India is ≈6.5%, which is much lower than that detected by our study. However, a more detailed analysis from the eastern India state of Odisha shows that this percentage can vary within a year, depending on season, and the highest reported prevalence of DENV/Plasmodium co-infections from this region was 31.8% during September–October, an observation similar to ours (28).

Although awareness of Plasmodium/DENV co-infections is increasing, little information is available about Plasmodium/Leptospira or Plasmodium/O. tsutsugamushi co-infections (13,29). This lack of information is concerning because our study suggests that Plasmodium/Leptospira co-infections are associated with severe malaria. Prevalence data for co-infections with these pathogens are limited. We emphasize the need for such information because although these pathogens are carried by different vectors, they co-exist in the same geoclimactic habitats that combine a warm, moist environment with dense vegetation and poor socio-economic development (13,29). The presence of one co-infecting pathogen can influence disease outcomes, treatment outcomes, development of immunity, or drug resistance with regard to infections caused by the other co-infecting pathogen. One example is the predisposition for bacteremia to develop in persons with malaria (30).

In most malaria-endemic settings, malaria is still diagnosed by microscopic examination of Giemsa-stained peripheral blood smears and rapid diagnostic testing for parasite antigen. The rapid test is specifically designed to detect P. falciparum and P. vivax and is extensively used because of its speed and simplicity. For microscopy, diagnostic success depends on the skill of the technicians who observe the peripheral blood smears. We found that rapid tests and microscopy missed most of the P. vivax–positive malaria cases and all P. knowlesi cases and detected only 1 of 18 Plasmodium mixed-species infections. This finding clearly shows the limitations of rapid testing and microscopy for comprehensive detection of malaria parasites, which have been independently observed in several studies and attributed to deletions in the HRP2 and HRP3 genes in the specific case of P. falciparum infection (31–34). This problem is well recognized for healthcare workers and researchers working toward malaria elimination all over the world. Although the rapid diagnostic test for malaria has been shown to be better than microscopic examination of Giemsa-stained peripheral blood smears, PCR has been shown to be far superior to rapid testing for diagnosing low-parasitemia malaria infections (35). Our observations were similar; PCR was most sensitive, followed by rapid testing and then microscopy. However, rapid tests have low success rates in areas of low transmission intensities and give rise to a high proportion of false negatives (36). In addition, rapid tests fail to detect infections with emerging pathogens, such as the simian parasite species P. knowlesi and P. cynomolgi, both known to infect humans (37). Although recent reports highlight the improvement of rapid tests for P. knowlesi detection by use of a cross-reacting pan-parasite lactate dehydrogenase feature, we were unable to detect P. knowlesi by using a pan-parasite lactate dehydrogenase–containing rapid test, possibly because of low parasitemia, below the detection limit of the rapid test (38). P. knowlesi, which was previously believed to be localized to Southeast Asia, has now been reported from various parts of the world as single case reports of travelers’ infections from areas including Oceania, Europe, and the Middle East (39–41). From India, P. knowlesi infection has been reported from the Andaman and Nicobar Islands in the context of drug resistance and in a recent study by our group in the context of acute kidney injury (26,42). Historically, P. knowlesi infection was discovered as a naturally occurring human infection in Malaysia in 1965 (43). The accurate diagnosis of P. knowlesi by use of PCR took ≈40 years from its initial discovery and gave a preliminary indication of the burden of this zoonosis in Sarawak, Malaysia (44). Until this point, P. knowlesi as a human infection was frequently misdiagnosed as P. vivax, P. malariae, or P. falciparum infection.

To assess whether the infections originated locally, we surveyed the P. knowlesi patients in our study group for the possibility of travel abroad or interaction with visitors from abroad within their family. The patients who responded to our questionnaire do not appear to have traveled abroad or to have had direct contact with anyone visiting them from abroad, suggesting local presence of P. knowlesi. However, unknown sources of travel from Southeast Asia, a neighbor to India, cannot be ruled out. Currently, testing for P. knowlesi is not included in diagnostic procedures in India, irrespective of diagnostic method (microscopy, rapid diagnostic test, or PCR), because it has not been widely reported. However, India is known to harbor both the potential vector for P. knowlesi, the Anopheles dirus mosquito, as well as the reservoir, pig-tailed macaques (Macaca nemestrina), thereby making India a potential ecosystem for the proliferation of this zoonotic Plasmodium species (45). Furthermore, the populations of Macaca mulatta macaques and related species have recently expanded in northern India, particularly in the state of Uttar Pradesh, which may explain the appearance of P. knowlesi in our study population representative of these areas, whereas it was not reported earlier (45). A recent report has also demonstrated the presence of P. falciparum parasites in monkey populations from India, indicating freely occurring undetected zoonotic transfer of the malaria parasites across reservoirs and hosts. Therefore, healthcare workers and national programs should incorporate all species of malaria parasites known to infect humans, in their diagnostic portfolio.

In conclusion, our study clearly showed that microscopy and rapid diagnostic testing gave false-negative results for most mixed-species infections and completely missed P. knowlesi infections, co-infections and mixed Plasmodium infections were highly prevalent in patients with malaria, Plasmodium/DENV co-infections were the most common co-occurring pathogens in our study population, P. knowlesi infections were present in India, Plasmodium/DENV4 co-infections were associated with mild malaria, and Plasmodium/Leptospira infections were associated with severe malaria. Although the ORs support the above findings, the 95% CIs for these associations were wide. CIs reflect the uncertainty of the estimated effect, and wider intervals suggest greater uncertainty. The wide 95% CIs in this study suggest that although trends were observed, additional data points are needed to determine the effect size of these associations. Wider prevalence studies investigating malaria co-infections are needed.

The government of India has recently declared a goal of malaria elimination by 2030, which will be a major step toward global malaria eradication because India serves as a major Plasmodium reservoir, contributing to almost 4% of malaria-related deaths globally. Therefore, to make malaria elimination possible, we offer 2 recommendations based on our observations in this study, particularly for tertiary healthcare centers or centers where the burden of severe malaria cases is high. First, malaria elimination efforts will need to include strategies for malaria elimination in humans as well as animal reservoirs. Second, efforts toward the development of novel diagnostics for malaria must be renewed, and AFI diagnoses must include screening for all 5 Plasmodium species, Leptospira, and all 4 DENV serotypes.

Supplementary methods and results for study of association of dengue virus and leptospira co-infections with malaria severity.