Retrospective study of association between Plasmodium falciparum malaria density and platelet count at Presbyterian Hospital, Agogo

Introduction

Malaria remains one of the most pressing global public health concerns, especially in sub-Saharan Africa, where the disease is both endemic and persistent (Oladipo et al., 2022). Among the different species of malaria parasites, Plasmodium falciparum stands out as the most dangerous and virulent, accounting for the highest number of malaria-related deaths globally (Yokota et al., 2025). It is a life-threatening infectious disease that is caused by protozoan parasites known as Plasmodium sp., which are transmitted from the female Anopheles mosquito to humans through its bite (Sato, 2021). It exerts a highly adverse impact particularly in children and pregnant women with maternal complications including anemia, spontaneous abortion, preterm delivery, and low birthweight driven by parasite placental sequestration (Osarfo et al., 2022). Globally, there are five classical species that cause human malaria: P. falciparum, P. malariae, P. ovale, P. vivax, and the zoonotic P. knowlesi (Sato, 2021). According to the World Malaria report 2024, there were approximately 263 million reported cases and 597,000 deaths globally, with 94% of cases and 95% of deaths occurring within the WHO African Region (Venkatesan, 2025). Transmission intensity remains high across Africa due to insecticide resistance, favorable climatic conditions and health system weakness (Venkatesan, 2024). Sub-Saharan Africa is particularly known to contribute to about 90% of the region’s total malaria burden (Adjei et al., 2023).

In Ghana, which ranks among the top ten highest malaria-burden countries globally, P. falciparum alone is responsible for 98% of all malaria cases, with the remainder attributable to P. malariae and P. ovale either as mono-infections or co-infections with P. falciparum (Aheto et al., 2024). Consequently, focusing exclusively on P. falciparum provides the most clinically relevant insights for target interventions in this region, as its high virulence and geographical dominance minimize the confounding variations in hematological parameters potentially introduced by mixed infections or less prevalent species. Malaria has been studied to impact various hematological parameters, significantly, namely red blood cell count, hemoglobin value, white blood cell and platelet counts (Sakzabre et al., 2020). The pathophysiology of P. falciparum involves parasite invasion, intraerythrocytic development and schizogony, leading to erythrocyte destruction and microvascular sequestration mediated by parasite-derived adhesion proteins such as PfEMP1 (Jensen et al., 2020; Zekar and Sharman, 2023).

Concurrently, malaria-associated thrombocytopenia has been widely documented, with growing evidence suggesting that the severity of thrombocytopenia correlates with peripheral parasite density (Ebraheem et al., 2025; Gebreweld et al., 2021; Kumar et al., 2022). Proposed mechanisms include immune-mediated destruction, splenic sequestration, increased platelet consumption, and impaired platelet production (Jiero and Pasaribu, 2021; Watson et al., 2022). However, this association is not consistently observed across all settings. Establishing this relationship is significant especially in low-resource setting, where platelet count may support malaria diagnosis and treatment monitoring.

While existing literature firmly established, the occurrence of thrombocytopenia in acute malaria, investigating the precise mathematical relationship between parasite density and platelet dynamics remains clinically vital. In resource-limited and rural settings where gold-standard microscopy or rapid diagnostic tests (RDTs) may face quality control challenges or temporary stock-outs, full blood counts are routinely available. Understanding this relationship could help clinicians utilize platelet trends as accessible adjunctive indicators to guide case management. However, a primary challenge in promoting platelet count is its low specificity; many non-malaria acute febrile illnesses prevalent in the tropics, such as dengue fever, typhoid fever, and viral infections simultaneously induce various degrees of thrombocytopenia. Thus, contextualizing how platelet variations parallel parasite burden is necessary to determine if they hold any true supportive or disease-stratifying value. Three notable gaps persist in the current literature: first, there is a very limited recent localized data in high-transmission pockets of Ghana mapping these specific parameters. Second, methodological variations across previous studies complicate the extraction of clear clinical trends. Third, many previous localized investigations have relied on relatively restricted sample sizes, which limits the statistical power required to characterize broad demographic and density-dependent trends (Khan et al., 2020; Mukkamala et al., 2025).

This retrospective observational study addresses these gaps by evaluating a substantial institutional dataset compiled over a 13-month period at the Presbyterian Hospital in Agogo, Ghana. Located in a holoendemic zone with Annual Entomological Inoculation Rate of 40–158 infective bites per person per year, this site provides an ideal epidemiological framework to clarify the real-world association between P. falciparum density and platelet variations across diverse demographic cohorts (Badu et al., 2013).

Materials and methods

Study setting

The study was conducted at the Presbyterian Hospital, Agogo-Ghana. This facility acts as a major district healthcare center, managing a high volume of annual malaria cases and utilizing integrated electronic medical databases for laboratory demographic tracking.

Study design

A retrospective observational framework was implemented to analyze historical laboratory records spanning from September 2024 to September, 2025. This approach captures established routine clinical patterns over a 13-month cycle without the necessity of active prospective patient tracking.

Study population

The study consisted of all patients for whom malaria test was requested at the selected hospital during the specified study period. This includes both male and female patients across all age groups. The hospital records provided detailed laboratory reports on malaria and platelet counts, which were the key variables for this study.

Inclusion criteria for the study were:

• Patient records with confirmed positive or negative for Plasmodium falciparum malaria by microscopy within the already-defined study period (September, 2024 to September,2025).
• Records with documented automated complete blood count platelet metrics.
• Laboratory entries with age and sex demographics clearly recorded

Exclusion criteria included:

• Records of individuals diagnosed with mixed malaria infections (e.g., P. vivax and P. falciparum).
• Data profiles with incomplete laboratory entries for any of the mandatory parameters required for this study.
• Records of individual for whom malaria test has been requested, but having chronic illnesses or conditions known to alter platelet counts (e.g., leukemia, HIV, or autoimmune disorders). These were identified by accessing their electronic medical records.
• Duplicate records or follow-up entries from the same patient during the same episode of illness.

Sampling technique

A retrospective census of all eligible records was conducted. All records meeting the predefined inclusion criteria within the study period were included in the analysis.

Sample size

This was determined based on the availability and completeness of medical records within the study period. All patient records that met the inclusion criteria were reviewed and a total population of 15,989 sets of data were obtained.

Data collection procedure

The data was collected from the hospital’s electronic records system; Lightwave Information Management System (LHIMS), with the assistance of the hospital’s records and laboratory departments. To protect patient confidentiality, all identifying information were anonymized.

Data collection tools

The study used a structured data extraction sheet for data collection unto Microsoft Excel. This instrument was designed to capture the key variables.

Data processing and analysis

The retrospective dataset underwent rigorous data cleaning and initial screening within Microsoft Excel. Duplicate entries and follow-up records from the same patient during a single episode of illness were systematically identified and removed to ensure that each data point represented an independent clinical encounter. Missing or incomplete values across the primary variables—age, gender, microscopic malaria parasite result, and automated platelet count—were filtered out prior to statistical execution. Subsequent statistical analyses were performed using R software (version 4.5.3). Data normality for continuous variables, specifically platelet counts and parasite densities, was evaluated using the Kolmogorov-Smirnov test for the total dataset. The test yielded a statistically significant deviation from normality (p = 0.001), justifying the exclusive use of non-parametric statistical methods. Descriptive statistics were compiled to characterize the overall analytical population. To evaluate the relationship between continuous variables at the individual patient level, Spearman’s rank correlation analysis was executed to compute the correlation coefficient (ρ) and its corresponding 95% confidence interval (CI). For categorical and stratified comparisons, such as evaluating platelet count distributions across distinct age cohorts and categorized parasite density groups; the non-parametric Kruskal-Wallis test was applied. Post hoc pairwise comparisons were subsequently conducted using Dunn’s test, with p-values adjusted via the Bonferroni correction to strictly control for Type I error inflation across multiple comparisons. For binary group comparisons, such as differences between malaria-positive and malaria-negative cohorts or between genders, the Mann-Whitney U test was utilized. To evaluate the clinical diagnostic capability of platelet count, a binary logistic regression model was constructed with Plasmodium falciparum infection status (positive vs. negative) as the outcome variable. The model’s overall performance was assessed using classification accuracy, the No-Information Rate (NIR) to account for class imbalance, sensitivity, specificity, and balanced accuracy. To determine the discriminatory power of platelet counts independent of a single arbitrary threshold, a Receiver Operating Characteristic (ROC) curve analysis was executed, and the Area Under the Curve (AUC) was calculated. The optimal probability threshold was determined using Youden’s J statistic to balance diagnostic sensitivity and specificity. Statistical significance for all analyses was established at a two-tailed threshold of p < 0.05.

Laboratory protocol for malaria slide preparation and parasite quantification

Routine malaria diagnosis and parasite quantification at the study site’s laboratory were performed by following standard protocols using Giemsa-stained thick and thin blood films. To ensure accuracy and minimize the propensity for over- or under-counting, thick films were utilized for lower parasite load, while thin films were employed were employed for higher loads in order to ease counting (Wynberg et al., 2023). Because this was a retrospective study utilizing documented data from routine clinical care, blood smears were evaluated once by the laboratory microscopists to guide immediate patient management, and hence independent research-specific double-reading of slides was not performed.

For quantification on thick blood films, trophozoites of P. falciparum parasites were counted against white blood cells (WBCs). If 100 or more parasites were observed after counting against 200 WBCs, the density was calculated immediately. If fewer than 100 parasites were observed after 200 WBCs, counting was continued up to 500 WBCs. Parasite density (parasites/µL) was calculated using each patient’s concurrent individual automated WBC count obtained from their complete blood count (CBC) results; standard assumed WBC count was not used (Wynberg et al., 2023).

Thick film formula:

Parasite Density(parasites/µL)=Number of Parasites CountedNumber of WBCs Counted×Patient′s WBC Count/µL

For the quantifications performed on thin film, the number of Parasitized red blood cells (RBCs) was counted against a minimum of 2,000 RBCs. The parasite density (per µL) was calculated using the individual automated RBC count (per µL) obtained from their CBC results (Alves and Carvalho, 2024).

Thin film formula:

Parasite Density(parasites/µL)=Number of Parasitized RBCsTotal RBCs Counted×Patient′s RBC Count/µL

Validity and reliability

The data extraction form was designed in alignment with established clinical indicators and standard hospital record templates. To ensure content validity, the instrument was reviewed by an expert panel specializing in clinical research and laboratory diagnostics, confirming that all primary and secondary study variables were accurately captured.

The research assistant underwent formal training to standardize the data extraction process and ensure consistency across all medical records. To minimize transcription errors, data entries were independently cross-checked by a secondary reviewer. Furthermore, the reliability of the source clinical data was supported by the use of standardized, routinely calibrated laboratory equipment; specifically, platelet measurements were performed using a Sysmex XN-1000 automated hematology analyzer (Sysmex Corporation, Kobe, Japan). Calibration was performed in accordance with the manufacturer’s recommendations and the laboratory’s standard operating procedures. Daily internal quality controls were carried out using commercial control materials at multiple levels (Level 1, Level 2 and Level 3) before patient sample analysis, and the laboratory participated in external quality assessment (EQA) program to ensure accuracy and reliability of platelet measurements. Routine preventive maintenance was also performed according to manufacturer’s recommendations. Malaria microscopy was conducted using a Zeiss Primo Star iLED microscope (Carl Zeiss AG, Oberkochen, Germany).

Ethical consideration

The study followed all ethical guidelines for research involving secondary data. Ethical approval was obtained from the Committee for Human Research, Publication and Ethics (CHRPE), Kwame Nkrumah University of Science and Technology (KNUST), and the Institutional Research Board of the Health facility before data collection began.

Results

Demographics

The study compiled data on malaria and platelet count tests from September, 2024 to September, 2025. After data cleaning, a total of 15,989 records were obtained. These represented patients who were tested for malaria and complete blood count simultaneously. The dataset constituted 69.50% females and 30.50% males, making it a female dominated dataset. The most dominant age group was young adult (25–44 years) which represented 32.05% of the population, the least represented population was senior (aged 75 years and above. Females in the 25 to 44 years age group were the most represented subgroup in the dataset representing 26.10% and males from age 75 years and above were the least represented subgroup (Table 1).

Table 1. Demographics by age and gender.

The bold values represent total sample used for the study.

Test positivity rate of Plasmodium falciparum malaria

A total of 15,989 blood films were prepared to be examined under the microscope for Plasmodium falciparum. After microscopic examination, 1,690 were reported to be positive for P. falciparum representing 10.6% of the entire dataset (Figure 1A).

Test positivity rate of P. falciparum by month

The positivity rate fluctuated across the months under study. June 2025 recorded the highest positivity rate (16.1%), followed by May 2025 (14.5%), while the lowest transmission rates were observed in February 2025 (5.5%) and September 2025 (5.6%) (Figure 1B).

Platelet count variations

Pairwise comparisons of platelet count among age groups of the entire study population

A Kruskal–Wallis rank-sum test showed a statistically significant difference in platelet counts across the six age groups (χ²(5) = 257.60, p < 0.001). The median platelet counts (interquartile range, IQR) varied across age categories, with the highest values observed in children (0–14 years) [239 (157–330)] and the lowest in seniors (≥75 years) [199 (151–247) × 10³/uL]. Intermediate median values were observed among youth (15–24 years) [227 (176–282) × 10³/uL], young adults (25–44 years) [210 (170–256) × 10³/uL], middle-aged adults (45–59 years) [208 (154–265) × 10³/uL], and the elderly (60–74 years) [208 (155–260) × 10³/uL] (Table 2).

Table 2. Kruskal–Wallis test comparing platelet count across age groups.

Post hoc analysis using Dunn’s test with Bonferroni correction indicated that children had significantly higher platelet counts compared with all other age groups (all adjusted p < 0.05), except that differences were not explicitly directional in the test output but are supported by the higher median values observed in children. Youth (15–24 years) also had significantly higher platelet counts compared with all older adult groups, including young adults, middle-aged adults, elderly adults, and seniors (all adjusted p < 0.001). In contrast, no statistically significant differences were observed among most adult groups aged 25 years and above, including comparisons between elderly and middle-aged adults (p = 1.000), elderly and young adults (p = 1.000), and middle-aged and young adults (p = 1.000). Differences between seniors and elderly (p = 0.183) and seniors and middle-aged adults (p = 0.097) were also not statistically significant after adjustment (Table 3).

Table 3. Dunn’s multiple comparisons test with Bonferroni adjustment for age groups.

Platelet count comparison among the entire population by gender

A Mann–Whitney U test revealed significantly higher platelet counts in females compared to males (W = 28,872,851, p < 0.001), with median values of [220 (IQR: 172–273) × 10³/uL] and [208 (IQR: 149–277) × 10³/uL] respectively (Table 4).

Table 4. Mann-Whitney U test comparison of baseline platelet counts stratifies by patient gender.

Platelet count variation among Plasmodium falciparum malaria patients by gender

The study also analyzed the average platelet count among males and females infected with Plasmodium falciparum. The average platelet count for males and females was 157.81 × 10³/uL and 157.79 × 10³/uL respectively. Comparing the averages, it is worth noting that the difference is not significant.

Platelet count comparison among P. falciparum malaria positive and negative cases

A Mann–Whitney U test (Wilcoxon rank-sum test with continuity correction) revealed a statistically significant difference in platelet counts between malaria status groups (Negative vs Positive) (W = 17,998,423, p < 0.001); Median platelet counts were 143 (IQR: 87–206) × 10³/uL for malaria-positive individuals compared to 223 (IQR: 175–279) × 10³/uL for malaria-negative individuals.

Platelet count distribution across P. falciparum parasite density categories and malaria negative individuals

The parasite density analyzed were categorized into low parasitemia(<1000 parasite/uL), moderate parasitemia (1000–9999 parasites/uL), high parasitemia (10,000–99999 parasites/uL) and severe parasitemia (≥100,000 parasite/uL) (Bohli et al., 2025). A non-parametric evaluation of platelet distributions revealed a progressive, stepwise decline in median platelet counts as Plasmodium falciparum parasite density increased from the uninfected baseline to severe parasitemia levels. The malaria-negative cohort exhibited the highest baseline platelet parameter, with a median count of 223 (IQR: 175–279) × 10³/uL. Among the infected individuals, patients with low parasitemia maintained the highest relative platelet values, showing a median of 177 (IQR: 116.5–235× 10³/uL. As density escalated to moderate levels, the median platelet count dropped noticeably to 131 (IQR: 83–190) × 10³/uL. This downward trajectory continued into the high parasite density group, which presented a median count of 106 (IQR: 64.8–161.2) × 10³/uL, and finally, the lowest absolute hematological profile was observed in the severe parasite density cohort, yielding a median platelet count of 91 (IQR: 60–139) × 10³/uL. Spearman rank correlation revealed a statistically significant, moderate negative correlation between Plasmodium falciparum parasite density and platelet count (ρ= -0.34, 95% CI: -0.39 to -0.30, p < 0.001) (Figure 1C).

Pairwise comparison of platelet count among parasite density categories and malaria negative individuals

A Kruskal–Wallis rank-sum test revealed a statistically significant difference in platelet counts across the five malaria parasite density groups (χ²(4) = 1232.93, p < 0.001). Subsequent Dunn’s post hoc test with Bonferroni adjustment demonstrated that most pairwise comparisons were statistically significant (Table 5).

Table 5. Kruskal–Wallis test comparing platelet count across parasite density groups.

Specifically, platelet counts differed significantly between the High and Low groups (p < 0.001), High and Moderate groups (p = 0.019), Low and Moderate groups (p < 0.001), High and Negative groups (p < 0.001), Low and Negative groups (p < 0.001), Moderate and Negative groups (p < 0.001), Low and Severe groups (p < 0.001), and Negative and Severe groups (p < 0.001). In contrast, there were no statistically significant differences between the High and Severe groups (p = 1.000) or between the Moderate and Severe groups (p = 0.333) after Bonferroni correction (Table 6).

Table 6. Dunn’s multiple comparisons test with Bonferroni adjustment.

Predictive performance of platelet count for malaria infection

A logistic regression model was developed to assess whether platelet count could predict malaria infection status (positive vs negative). The model demonstrated an overall accuracy of 89.9% (95% CI: 88.9–90.7%) alongside a no-information rate of 89.4%. The model yielded a balanced accuracy of 0.55, with sensitivity of 0.99 for detecting malaria-negative cases, and a specificity of 0.10 for malaria-positive cases. Receiver operating characteristic (ROC) analysis further showed that the model had an area under the curve (AUC) of 0.771. At the optimal probability threshold of 0.133, the model achieved a sensitivity of 63.6% and specificity of 79.9%. Across all statistical approaches, platelet count demonstrated a consistent and strong association with malaria infection and parasite burden. Individuals with malaria infection exhibited significantly lower platelet counts compared to uninfected individuals, and platelet levels decreased progressively with increasing parasite density. These findings were robust across univariate (non-parametric tests), correlation analysis, and multivariable regression models. The logistic regression model showed acceptable discrimination (AUC = 0.771), but poor sensitivity for detecting malaria-positive cases at standard classification thresholds.

Discussion

Demographics and temporal dynamics of P. falciparum

This study revealed a distinct demographic distribution, with females constituting 69.50% of the institutional population visiting the hospital for concurrent malaria and complete blood count testing. This imbalance is consistent with health-seeking patterns documented elsewhere in Ghana and broader European cohorts, where women display a higher frequency of healthcare utilization (Dakorah et al., 2022; Golinelli et al., 2025). This pattern is frequently shaped by reproductive health needs, maternal care, and cultural variations in health literacy and stoicism between genders (Höhn et al., 2020; Rata Mohan et al., 2025). While this female dominance reflects institutional attendance patterns, at the Presbyterian Hospital, Agogo, it represents a potential source of selection bias that may limit the direct generalizability of the crude infection rates to the wider community. However, downstream stratified analyses demonstrated that the underlying biological relationship between P. falciparum presence and platelet dynamics did not differ significantly between genders, suggesting that this demographic skew does not invalidate the hematological associations observed.

The observed overall P. falciparum slide positivity rate of 10.6% aligns closely with recent cross-sectional data from all 29 districts of the Greater Accra Region of Ghana (0.9-10% and 1.4 - 15%) and multi-country sub-Saharan observations (6.4% to 36.4%) (Dosoo et al., 2024; Tembo et al., 2024). However, institutional positivity rate is lower than community-based studies in highly endemic transmission pockets, such as southwest Ethiopia (50.72%) (Alemu et al., 2024). The comparatively lower rate may reflect the success of targeted public health interventions within the Agogo township and its environs, including enhanced compliance with Intermittent Preventive Treatment in pregnancy (IPTp) and widespread distribution of insecticide-treated nets. Temporally, malaria test positivity rate exhibited seasonal fluctuations, peaking sharply in May (14.53%) and June (16.07%). While precise local meteorological variables (such as exact rainfall volume and relative humidity values) were not directly modelled in our dataset, requiring caution in drawing definitive environmental conclusions. These peaks coincide historically with the major rainy seasons in Ghana, spanning April to July (Owusu and Waylen, 2013). The observed surge in cases during these months likely reflects the expanded availability of breeding sites for Anopheles vector, accelerating transmission dynamics within the population.

Platelet count variations by age and gender

An inverse descriptive trend was observed between baseline platelet counts and advancing age across the entire study population, with pediatric patients (0–14 years) exhibiting the highest median platelet counts [239 (IQR:157–330) ×10³/µL] and seniors exhibiting the lowest [199 (IQR: 151–247) ×10³/µL]. This general decline is consistent with observations in both malaria-endemic and non-endemic cohorts (Kumar et al., 2022; Kunapaisal et al., 2023). Multiple physiological factors account for this age-related reduction independent of acute infection, including a natural senescent decline in bone marrow activity, hematopoietic stem cell exhaustion, subclinical bone marrow fibrosis, and the biochemical effects of concurrent chronic medications (Cheng et al., 2025; Zou et al., 2025). Consequently, clinicians must account for age-stratified baseline shifts when evaluating the magnitude of acute thrombocytopenia during febrile episodes. In contrast, when isolating the malaria-positive cohort, the average platelet counts between infected males (157.81×10³/µL) and infected females (157.79×10³/µL) were nearly identical, displaying no statistically or clinically significant variance. This finding strongly indicates that while gender influences baseline healthcare-seeking behavior and institutional demographics, it exerts no significant biological influence on the severity or manifestations of malaria-induced platelet depletion once infection is established, a conclusion corroborated by similar species-specific studies in other endemic regions (Saleem et al., 2024; Shah et al., 2023).

Impact of P. falciparum infection and parasite density on platelet dynamics

The investigation revealed a profound, statistically significant reduction in median platelet counts among P. falciparum-positive individuals [143 (IQR: 87–206) ×10³/µL compared to their malaria-negative counterparts [223 (IQR: 175–279) ×10³/µL (p < 0.001). This marked hematological divergence confirms that acute thrombocytopenia is a hallmark pathological feature of falciparum malaria, validating previous comparative studies conducted globally (Elkhalifa et al., 2021; Gupta et al., 2019). The primary underlying mechanisms driving this acute reduction include immune-mediated platelet destruction triggered by parasite-specific antibodies, accelerated peripheral consumption within microvascular beds, and acute splenic sequestration (Gurumurthy et al., 2026). When examining the parasite density cohorts, a clear progressive decline in median platelet counts was observed from the malaria negative group, through to the severe parasitemia group (i.e. malaria negative group: [223 (IQR: 175–279) × 10³/uL], low parasitemia: [177 (IQR: 116.5–235× 10³/uL], moderate parasitemia: [131 (IQR: 83–190) × 10³/uL], high parasitemia: [106 (IQR: 64.8–161.2) × 10³/uL], severe parasitemia: [91 (IQR: 60–139) × 10³/uL]. Spearman rank correlation confirmed a statistically significant negative correlation between individual P. falciparum parasite density and platelet count (ρ= -0.34, 95% CI: -0.39 to -0.30, p < 0.001). This inverse relationship indicates a direct biological link between peripheral parasite burden and the velocity of platelet clearance or sequestration, matching findings from several recent investigations. However, post hoc pairwise analysis using Dunn’s test with Bonferroni correction revealed a critical nuance: while platelet counts differed significantly between low, moderate, and high parasitemia groups, there was no statistically significant difference between the high and severe parasitemia cohorts (p = 1.000), nor between the moderate and severe cohorts (p = 0.333). This statistical overlap and subsequent plateau effect suggest that platelet depletion dynamics may not follow a strictly linear trajectory at higher parasite burdens. Instead, the pathological mechanisms driving consumption and sequestration appear to reach a point of physiological saturation, or are counterbalanced by acute compensatory bone marrow release, preventing infinite, linear platelet declines even as peripheral parasitemia escalates exponentially.

Clinical utility and predictive performance of platelet counts

Because this study demonstrates a highly significant statistical relationship between infection status, rising parasite density, and falling platelet counts across multiple univariate and correlational frameworks, evaluating its actual mathematical predictive capability is vital for case management. The built logistic regression model yielded a deceptively high raw classification accuracy of 89.9% (95% CI: 88.9–90.7%). Crucially, multi-variable interpretation reveals that this metric is heavily confounded by the underlying class imbalance of the population, as the No-information rate (NIR) was 89.4%. When adjusting for this skew, the model displayed a poor balanced accuracy of 0.55, indicating that its standalone, uncalibrated capability to differentiate true positives from negatives is severely restricted (sensitivity = 0.10, specificity = 0.99). To overcome this majority-class bias, Receiver Operating Characteristic (ROC) analysis was applied, mapping an Area Under the Curve (AUC) of 0.771. This represents an acceptable, intermediate biological discrimination capacity that fits comfortably alongside established diagnostic modeling frameworks in the literature. For instance, diagnostic frameworks utilizing multivariable logistic regression and ROC curve modeling automated blood counts (specifically emphasizing the sharp diagnostic weight of platelet parameters) have mapped highly discriminative AUC trajectories reaching 0.877 (Kotepui et al., 2015). By optimizing our specific decision threshold to 0.133 using Youden’s J statistic, our diagnostic model successfully raised sensitivity to 63.6% alongside a specificity of 79.9%. This localized performance profile reinforces broader facility-based ROC mapping trends (Algın et al., 2023) demonstrating that while automated platelet indices offer excellent objective visibility for acute tracking, their clinical utility rests in screening and risk stratification rather than operating as definitive standalone classifiers.

Study limitations

This study has several limitations that must be acknowledged:

• Retrospective design: The reliance on historical databases prevented active clinical follow-up or confirmation of patient treatment outcomes.
• Lack of clinical severity metrics: The dataset lacked comprehensive secondary clinical records (i.e. presence of organ dysfunction, cerebral symptoms, or severe anemia), preventing an evaluation of whether the observed thrombocytopenia correlated directly with severe clinical malaria presentations rather than purely high parasitemia
• Residual confounding: While major chronic comorbidities (such as leukemia, HIV, and documented autoimmune disorders) were actively excluded by checking electronic medical records, we could not account for transient acute co-infections (e.g., arboviruses, bacterial infections), localized nutritional deficiencies (such as folate or Vitamin B12 deficiency), or the use of over-the-counter medications prior to hospital presentation.
• Single-reading microscopy: Because data was derived from real-time routine care rather than a prospective trial framework, independent research-specific double-reading of blood smears was not performed, introducing potential operator-dependent variation in parasite quantification.
• Institutional selection bias: The dataset exhibited a pronounced female dominance (69.50%). While stratified analyses confirmed that gender did not biologically alter malaria-induced platelet depletion, this demographic skew requires caution when attempting to generalize the crude infection rates directly to the broader community or general population.

Conclusions

Based on the findings from this study, it is concluded that Plasmodium falciparum malaria remains a notable public health concern among patients seeking healthcare at the Presbyterian Hospital, Agogo, presenting an institutional positivity rate of 10.6%. The significantly lower median platelet counts observed in P. falciparum-positive patients compared to their malaria-negative counterparts confirm that acute thrombocytopenia is a consistent and pathophysiologically grounded hallmark of this parasitic infection. Age-stratified analyses revealed a clear baseline decline in platelet counts with advancing age across the broader population, and infected older cohorts (aged 45 and above) demonstrated lower absolute platelet counts, indicating that older patients are more vulnerable to severe manifestations of thrombocytopenia during acute malaria episodes. Conversely, gender does not exert any significant biological influence on the severity of malaria-induced platelet depletion, despite driving prominent variations in baseline healthcare-seeking behaviors and hospital attendance demographics. Crucially, while non-parametric testing and correlation analysis (ρ = -0.34, p < 0.001) supports a biologically plausible association between escalating peripheral parasite density and falling platelet levels, predictive modeling demonstrates that platelet counts possess only moderate diagnostic classification performance (AUC = 0.771). Due to extensive distribution overlaps and class imbalance limitations, a standalone platelet measurement exhibits insufficient sensitivity and specificity for definitive diagnostic categorization. Therefore, thrombocytopenia cannot substitute for direct parasitological verification. It should, however, be systematically integrated into clinical practice as a highly accessible, rapid, and valuable supportive laboratory indicator alongside gold-standard microscopy to optimize patient triage, flag high-probability cases, and support risk stratification in resource-limited diagnostic settings.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Chim Chan, Osaka Metropolitan University, Japan

Reviewed by: Adilson José DePINA, CCS-SIDA/MoH, Cabo Verde

Diana Shintawati Purwanto, Universitas Sam Ratulangi Fakultas Kedokteran, Indonesia