Development and validation of a risk prediction model for pulmonary tuberculosis in presumptive tuberculosis patients in Tigray, northern Ethiopia – Nature.com 20:16  

Scientific Reports volume 15, Article number: 32270 (2025) Cite this article

Abstract

The incidence of tuberculosis (TB) has increased in Tigray, Ethiopia due to war and a crippled healthcare system. Although early detection and treatment are critical for TB control, over 30% of TB cases are missed using current diagnostic techniques. Thus, we developed and validated a risk prediction model for pulmonary TB in presumptive cases. In this multicenter cross-sectional study, we consecutively enrolled 907 respondents from primary healthcare facilities in Tigray, northern Ethiopia. We used least absolute shrinkage and selection operator regression to identify variables for the model. Risk scores were generated from the coefficients of multivariable logistic regression. We evaluated the model performance using the area under the curve and calibration plots, and clinical utility using decision curves. Among all respondents, 155 (17%) had GeneXpert-confirmed pulmonary TB. At an optimal cutoff value of 8.5, the model demonstrated a discrimination accuracy of 0.82 (95% CI: 0.78–0.85), a sensitivity of 82.6%, and a specificity of 68.9%. The model had a calibration slope of 0.98 and an intercept of 0.001. The model exhibits acceptable discrimination and calibration performance. Thus, it can be used for screening patients for pulmonary TB in primary healthcare settings where accurate diagnostic resources are limited.

Similar content being viewed by others

The effect of empirical and laboratory-confirmed tuberculosis on treatment outcomes

Article Open access21 July 2021

Prevalence of extrapulmonary tuberculosis and factors influencing successful treatment outcomes among notified cases in South India

Article Open access10 March 2025

Magnitude of extrapulmonary tuberculosis and its associated factors among TB suspected patients at Wolkite University specialized hospital in central Ethiopia

Article Open access21 May 2025

Introduction

Tuberculosis (TB) is a major public health problem worldwide, with 10.8 million new cases and 1.25 million deaths in 2023 1. More than 95% of TB deaths occur in low- and middle-income countries2. Ethiopia has one of the highest rates of TB, with 146 cases per 100,000 population in 2023 1. Between 2010 and 2020, the country had decreased TB incidence rate by an average of 5% yearly3. Nevertheless, efforts to combat the disease have been disrupted due to the recent civil war in certain regions of the country4. In Tigray, northern Ethiopia, TB remains a major threat during the prewar, active war (2020–2022), and postwar periods. According to the Global Burden of Disease study, 178 TB cases per 100,000 individuals were reported in Tigray in 2019 5. The war led to a threefold increase in TB incidence in the region in 2024 4.

Early detection and treatment of every patient with the disease is fundamental for TB control7. Like any other patient, TB suspects in Tigray, northern Ethiopia visit primary healthcare (PHC) facilities for essential services. These facilities are staffed by mid-level healthcare providers (MLHPs)8. MLHPs can diagnose and treat TB patients in the presence of appropriate diagnostic tools, particularly automated machines that perform most of the steps. However, most PHC facilities have limited TB diagnostic resources and have been severely affected by the Tigray War9. Although the GeneXpert MTB/RIF test is reported to be the most cost-effective diagnostic tool for Ethiopia, its availability in PHC facilities is limited10. Therefore, suspected cases of TB are diagnosed using smear microscopy, which misdiagnoses half of them because of its limited sensitivity11.

In Ethiopia, more than 30% of patients with drug-susceptible TB and 70% of patients with multidrug-resistant TB remain undetected12. The problem could be even worse in Tigray, where the war destroyed the majority of health facilities and ruined healthcare providers, resulting in fatalities, displacement, joblessness, and substantial stress at work13,14,15. Therefore, care providers in this setting should be supported by assistive tools in decision-making. For instance, a risk prediction model can support MLHPs to identify individuals who are more likely to have pulmonary TB (PTB), enabling prompt diagnosis confirmation and anti-TB medication planning16.

The PHC unit brings care closer to the community, making it an ideal level for intervention to reach missed TB cases. In addition, evidence shows that if MLHPs are adequately trained, supported, and supervised, they can provide TB care of the desired quality17. Thus, designing a decision-support tool for MLHPs to aid in the diagnosis of TB in this setup is necessary. To this end, considering contextual circumstances and care providers’ preferences in parameter selection is crucial18. In Ethiopia, few studies have developed risk prediction models to predict TB among HIV patients19 and the community20. However, none of these studies considered the preferences of healthcare professionals. As Tigray from northern Ethiopia continues to experience a high burden of TB, developing a prediction model would help the MLHPs identify PTB cases. Therefore, in this study we developed and validated a risk predictive model for PTB in presumptive cases of TB in the PHC settings of Tigray, northern Ethiopia.

Methods

Study design and setting

This multicenter cross-sectional study was conducted in the Tigray region of Ethiopia from October 2023 to June 2024. Tigray is Ethiopia’s northernmost region, which was hit by the civil war from November 04, 2020, to November 03, 202221. The region comprises seven zones, four of which (Eastern, Mekelle, Southeastern, and Southern zones) were purposively included in the study. These four zones have the highest rates of TB morbidity and mortality in the region. The study was conducted in 38 public health facilities (10 primary hospitals and 28 randomly selected health centers) in PHC settings. A sputum smear examination is performed in most PHC facilities to diagnose TB. All general and referral hospitals offer a GeneXpert MTB/RIF test. During the study period, only five primary hospitals in the study setting were performing rapid molecular tests. However, there was a system for TB specimen referral and result delivery among nearby health facilities.

Study population and eligibility

We included presumptive TB cases among those aged 15 years or older in primary care. Patients receiving anti-TB treatment and those undergoing follow-up were excluded.

Sample size and sampling technique

A total of 907 patients were recruited from public health facilities. The sample size was calculated based on the methodology described by Riley et al. for the multivariable prediction model of binary outcomes22. Patients suspected of having TB were consecutively enrolled.

Data collection

We collected sociodemographic, clinical, and behavioral data using a structured, interviewer-administered questionnaire at the outpatient department of PHC facilities. Predictor information was obtained from patients before GeneXpert testing. Predictor assessment and diagnostic confirmation were performed in different departments or institutions. Data were collected by trained healthcare workers.

Diagnosis of PTB was confirmed bacteriologically using the GeneXpert test. This test has a sensitivity of 88% and a specificity of 98% for identifying PTB23. PTB status was determined two hours following the predictor’s assessment provided the test was performed in the same facility or after five working days if the GeneXpert test was performed at a different institution. All laboratory investigations were conducted as part of the standard routine care.

Data processing and analysis

The data were entered into Kobo Collect v2023 1.2 and exported to Excel for data cleaning. We checked the data for errors and assigned non-overlapping numerical codes for separate answers in the questionnaire. The data were exported to R statistical software (version 4.4) for analysis. Missing data were detected for age (2.4%), family size (1.5%), number of rooms (1.5%), weight (4.5%), and height (7.6%). These missing values were managed by multiple imputations via the ‘mice’ package in the R software. We used predictive mean matching to impute numeric variables. We checked the assumptions of the chi-square test for categorical variables and normality for the continuous variables. Multicollinearity among the predictors was checked using the variance inflation factor (VIF), and all the VIF values were < 5, indicating that multicollinearity was not a problem.

Variable selection

We identified candidate predictors based on their strength of association and clinician recommendations from our previous studies18,24. Twenty-seven variables with a p-value < 0.2 in the bivariable binary logistic regression were fitted into the least absolute shrinkage and selection operator (LASSO) regression. LASSO is a regularization and feature selection method that applies a penalty to minimize overfitting and enhance model accuracy25. For the LASSO logistic regression, we split our data into training (80%) and testing (20%) sets. We then fitted our model through the ‘glmnet’ package of R software to the training data using 10-fold cross-validation. We set the tuning parameter lambda to a minimum and one standard error (1se) to shrink the coefficients and select the variables. Finally, we evaluated the performance of the LASSO on the testing dataset using an accuracy parameters.

Model development

Variables identified by the LASSO method were entered into a multivariable binary logistic regression model. We predicted the probability of PTB based on ten predictors. We presented an adjusted odds ratio (AOR) and 95% confidence interval (CI) to show the association between selected predictors and the outcome. In the final model, a p-value of < 0.05 was considered statistically significant.

Risk score derivation and nomogram construction

We computed a risk score by assigning points to the variables identified as significant predictors in the final model. Weights were generated from the multivariable regression model by dividing all coefficients by the smallest beta coefficient. We then rounded it to the nearest integer to ease the calculation of the score by frontline healthcare workers. We used Youden’s index method to determine the optimal threshold to classify patients as “high risk” or “low risk.” In addition, we created a nomogram to visualize the predictors and their contributions in a single plot. The nomogram is a graphical notation that can be easily interpreted without manual calculation.

Model performance (discrimination and calibration)

A receiver operating characteristic (ROC) curve was used to assess the model’s classification performance. We evaluated the discrimination capacity of the model using the area under the ROC curve (AUC) and calibration using a calibration plot with its slope and intercept. The sensitivity, specificity, and positive and negative predictive values of the risk scores were calculated.

Model validation

External validation was not performed because of budgetary constraints. However, we used a bootstrap method to validate our model internally. Bootstrapping is a resampling method with a replacement that reflects the drawing of samples from an underlying population. The bootstrap samples were of the same size as the original sample. In the current internal validation, we performed 200 bootstrap samples, which were sufficient to obtain stable estimates26.

Clinical utility

We evaluated the clinical utility of the prediction model using decision curve analysis (DCA). DCA determines whether using the model to guide clinical decisions improves outcomes compared with alternative strategies, such as treating all or none27. DCA focuses on the net benefit (NB) of a model across a range of probability thresholds, balancing the true and false positives.

To sum up, this study was reported according to the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) guidelines for developing, validating, or updating a prediction model28.

Ethical consideration

The Institutional Review Board (IRB) of the College of Health Sciences, Mekelle University, granted ethical approval (MU-IRB2013/2022). A letter of support was obtained from the Tigray Regional Health Bureau, Ethiopia to conduct the study. The nature and goals of the investigation were explained to all participants. Informed consent was obtained from all participants and/or their legal guardian(s). All study records were stored safely. Furthermore, by collectively analyzing and sharing the data, the study participants’ privacy and anonymity were guaranteed. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Results

Socio-demographic and behavioral characteristics of the study subjects

Overall, 907 presumptive TB cases, aged ≥15 years were prospectively enrolled in the current study. The mean age of the study participants was 43.6 ± 17.7 years. Among the participants, 479 (52.8%) were men, 494 (54.5%) lived in rural areas, and 371 (40.9%) were unable to read and write. One hundred and fourteen (12.6%) patients had a history of TB contact, and 130 (14.3%) had a history of imprisonment (Table 1).Table 1 Sociodemographic and behavioral characteristics of the study subjects by their GeneXpert results.

Full size table

Clinical characteristics of the study subjects

Among the 907 respondents, 787 (86.8%) reported cough for at least 14 days, 559 (61.6%) reported fever, 603 (66.5%) reported weight loss, and 658 (72.5%) reported night sweats (Table 2). Additionally, 725 (79.9%) patients experienced fatigue, 160 (17.6%) had pallor, and 251 (27.7%) had crepitation. Among all presumptive TB cases, 155 (17.1%) were confirmed to have PTB.Table 2 Clinical characteristics of the study subjects by their GeneXpert results.

Full size table

Five hundred and sixty-five (62.3%) patients received antibiotics before the sputum smear test, of whom 20% (117) had PTB. In total, 138 (15.2%) study participants were positive for human immunodeficiency virus (HIV), and 17 (1.9%) had diabetes mellitus. However, 57.3% and 59.8% of the participants were unaware of their HIV and diabetes mellitus statuses, respectively (Table 3).Table 3 Selected risk factors and hematology profiles of the study subjects by their GeneXpert results.

Full size table

Variable selection

We used LASSO regression to select candidate variables. Initially, we fitted the LASSO logistic regression with 27 variables (sociodemographic, clinical, and risk factors). Next, we used a minimum lambda to shrink the coefficients (Fig. 1). Finally, we reduced the number of features to ten using one standard error value of lambda (lambda.1se) and produced a parsimonious and interpretable model. Our cross-validation yielded an accuracy of 83.4%. The minimum lambda (λ) value was 0.0046, with a log (λ) value of -5.37.

figure 1
Fig. 1

Model development

Ten variables were fitted in the multivariable binary logistic regression model. This included age, cough severity, loss of appetite, number of classical TB symptoms, antibiotic trial, history of TB contact, history of imprisonment, chronically sick-looking, pallor, and presence of dull sounds (Table 4). In the adjusted analysis, all variables, except for history of imprisonment, were independently associated with PTB (p < 0.05).Table 4 Association of key variables with GeneXpert confirmed PTB.

Full size table

Risk score derivation

The predicted risk of PTB in presumptive TB cases was calculated with the following formula: predicted risk =[Math Processing Error]

[Math Processing Error]

. Where ‘e’ is the base of natural logarithms and Z is the sum of scores, including the intercept.

To convert the raw score into a user-friendly scale, we used a simple linear scaling:

[Math Processing Error]

Where, [Math Processing Error] is the raw log-odds score, which is equal to Z in the preceding equation; ‘K’ is a scaling factor, and ‘C’ is a constant to shift the score. In our risk score, we assigned 1 for K and 5.62 for B. The possible sum of the scores for an individual in the dataset ranged from zero to 15. The best cutoff score (threshold) for our model was 8.5. This was obtained at a Youden index of 51.5%.

Nomogram for pulmonary TB

This nomogram calculates the risk of PTB in presumptive TB cases based on significant predictors. In this nomogram, once a care provider determines the patient-specific parameters, the patient’s total points are determined by adding the points earned for the values of each variable. Then, the probability of PTB in the patient can be ascertained from the total points (Fig. 2).

figure 2
Fig. 2

Model performance (discrimination and calibration)

The discrimination power (AUC) of the model for predicting PTB was 0.835 (95% CI: 0.80–0.87) (Fig. 3A). In the calibration plot, the predicted risks overlapped with the observed proportion of PTB (Fig. 3B). Our model had a slope of 0.98 (95% CI: 0.83, 1.17) and an intercept of 0.001 (95% CI: -0.02, 0.02). At this threshold value (≥ 8.5), the risk score showed moderate discrimination, with an AUC of 0.82 (95% CI: 0.78–0.85).

figure 3
Fig. 3

In addition, at the threshold score≥8.5 and a Youden’s index of 51.5%, our risk score had a sensitivity of 82.6% (95% CI, 75.7–88.2%) and a specificity of 68.9% (95% CI, 65.4–72.2) (Table 5). According to the cutoff score, 39.9% (362) of the study participants were at high risk (score ≥8.5), and 60.1% (545) were at low risk (score < 8.5). The proportions of patients with PTB were 35.4% and 5.0% among high-risk and low-risk patients, respectively.Table 5 Performance of the risk prediction model at classification scores < and ≥ 8.5 using GeneXpert test as a reference test (N=907).

Full size table

Model validation (Internal)

We performed bootstrap validation and determined the optimism-corrected performance of the model. The AUC for our model was 0.835 (95% CI, 0.80–0.871) after resampling. The internal validation had a mean absolute error of 0.014 and a mean squared error of 0.00027. The bias-corrected calibration had an intercept of -0.1025 and a slope of 0.905 (Supplementary Fig. S1). The Somer’s delta (Somer’s D) value for the resampled data was 0.636, which was approximately equal to that of the original data (Somer’s D = 0.67). In addition, the model had an optimism coefficient of 0.0339, indicating an internally valid prediction.

Clinical utility

According to the DCA, the curve for the model was higher and further to the right, indicating greater benefits of the model across a wide range of threshold values. The optimal threshold that maximized the net benefit was between 0.1 and 0.2. In this range, the model curve lies above the treat-all or treat-none lines. Thus, treating high-risk patients identified by the model can lead to favorable outcomes (Fig. 4).

figure 4
Fig. 4

Discussion

In the present study, we prospectively developed a risk prediction model to help MLHPs identify PTB among presumptive TB cases in PHC settings. The model was derived using predictors previously identified as important, straightforward, reliable, generalizable, and desirable by care providers in the study area18.

The performance of a model relies on the power of its constituents because, when combined, they produce an accurate risk prediction rule29. Our model was derived from the risk factors, signs, and symptoms of TB that care providers routinely collect. For example, a history of TB contact and age were significant risk factors included in the model; specifically, adolescents and young adults were at a greater risk of developing TB compared to others. Similarly, a previous study in China reported that patients aged 15–24 and 25–34 years had the greatest burden of TB30. This indicates that the disease primarily affects individuals in their most productive years1. Ethiopia is one of the 30 nations with a high burden of TB and HIV31. Between 30% and 63% of new HIV infections occur in men between the ages of 20 and 29 years and women between the ages of 15 and 24 years32. This strengthens the reliability of our risk prediction model.

Weight loss, cough, fever, and night sweats are classic symptoms of TB33. Patients with more than two classical symptoms have a 3.5-fold higher chance of developing the disease. Consequently, the largest weight was assigned to this variable, with a score of 3. This finding is lower than that of Baik et al. (2020) who reported that patients with three classical TB symptoms had a 5.4-fold higher chance of PTB than those who had only one symptom34. In the same study, as the number of classical symptoms increased to four, the risk of PTB increased by 10-fold, higher than that in our study. A possible explanation for this might be the difference in the scale of the variable, study design (case-control vs. cross-sectional), and sample size (1387 vs. 907).

Cough is a typical symptom of PTB and is a mode of transmission35; it is used as a screening method to initiate TB testing or track treatment response. A frequent, intense, and disruptive cough is categorized as a severe form36. In this study, patients with a severe cough had an 86% increased risk of developing PTB. Similarly, other studies reported positive findings in acoustic epidemiology, attempted to objectively measure cough patterns, and suggested it as a possible biomarker for detecting PTB37. The results also indicated that a history of TB contact was a strong predictor in the model. Other evidence also revealed that patients who had close contact with TB patients have a greater chance of acquiring Mycobacterium TB and subsequently developing active disease38.

In our model, the antibiotic trial was an important diagnostic factor for PTB. In support of this, an earlier study noted that a lack of response to antibiotics was a strong predictor of TB39. However, evidence from a systematic review argued against the benefit of antibiotic trials to rule out TB among smear-negative patients and urged care providers to discontinue this practice40. This difference may be attributed to the timing of antibiotic administration. In the study setting, broad-spectrum antibiotics were routinely prescribed to patients with TB symptoms before sputum tests were performed. This assumes that non-response to antibiotic treatment is an exemption to other differential diagnoses and that MLHPs empirically think about TB. Therefore, a history of antibiotic use should be proactively considered in risk prediction models for PTB.

In the present study, pallor, a chronically sick-looking appearance, and dull sound were significant predictors and were included in the risk-prediction model. Although pulmonary signs are nonspecific, they can indicate disease chronicity41. Delayed TB diagnosis is prevalent in the study area, with a median patient delay of 30 days42. This suggests that patients usually arrive at health facilities when they are emaciated or debilitated. By then, the chronic manifestations of TB become evident and receive the attention of the MLHPs. Many prediction models for PTB include chest X-ray findings, HIV infection, and the erythrocyte sedimentation rate and have improved diagnostic performance20,34,43,44. However, because most health facilities lack the necessary tools and supplies needed to conduct these examinations, it was not possible to gather data for all presumptive cases.

When a blend of significant predictors was used, the discriminatory performance of our simplified risk score was good (AUC = 82.0%). It also had a sensitivity and specificity of 82.6% and 68.9%, respectively. According to World Health Organization guidelines, a TB screening algorithm should have a minimum sensitivity and specificity of 90% and 70%, respectively45. However, the performance requirements for new TB tests have been relaxed because of the pressing need to enhance the case detection rate of TB. Some stakeholders demand high sensitivity, whereas others require only minor adjustments if the test is accessible to a broader community46. Therefore, our model can be applied in PHC facilities that lack rapid molecular tests to help MLHPs make informed decisions, reduce patient waiting times, and lower costs. Additionally, the model can be used to determine which patients should undergo more expensive diagnostic procedures, including chest radiography, mycobacterial culture, and other molecular tests47. However, external validation is needed to evaluate its applicability to various TB epidemiology.

Our study has the following limitations. First, it has not been externally validated and is prone to inter-observer variation during physical examination; therefore, the model’s performance could be affected. Second, the use of a GeneXpert rather than a culture test (gold standard) to confirm the outcome might have resulted in a misclassification of a few PTB patients. Finally, the model was derived from data of patients who sought care for TB symptoms; therefore, it cannot be used for TB screening in a community.

In conclusion, we developed a risk prediction model for PTB in presumptive TB cases at the primary healthcare level. The model had acceptable discriminating performance. Therefore, it can be used for TB screening or triaging in a clinical setup. Furthermore, the model can enhance TB case detection and reduce diagnostic delays.

Data availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

References

  1. WHO. Global Tuberculosis Report 2024 (World Health Organization, 2024).
  2. PIN. TB Is the Deadliest Infectious Disease. So why Haven’t You Heard of It? (Partners In Health, 2024).
  3. Agizew, T. B. et al. Prospects for tuberculosis elimination in ethiopia: feasibility, challenges, and opportunities. Pan Afr. Med. J. 43, 146. https://doi.org/10.11604/pamj.2022.43.146.35557 (2022).Article PubMed PubMed Central Google Scholar 
  4. Gebrehiwot, K. G. et al. War-related disruption of clinical tuberculosis services in tigray, Ethiopia during the recent regional conflict: a mixed sequential method study. BMC Confl. Health. 18, 29 (2024).Google Scholar 
  5. IHME. VizHub – GBD Results from 1990 to 2019., https://vizhub.healthdata.org/gbd-results/ (2023).
  6. Gebregergs, G. B., Berhe, G., Gebrehiwot, K. G. & Mulugeta, A. Predicting tuberculosis incidence and its trend in tigray, ethiopia: A Reality-Counterfactual modeling approach. Infect. Drug Resist. 17, 3241–3251. https://doi.org/10.2147/idr.s464787 (2024).Article PubMed PubMed Central Google Scholar 
  7. WHO. The End TB Strategy (World Health Organization, 2014).
  8. WHO. Primary Health Care Systems (PRIMASYS): Case Study from Ethiopia, Abridged Version (World Health Organization, 2017).
  9. Gesesew, H. et al. The impact of war on the health system of the Tigray region in ethiopia: an assessment. BMJ Global Health. 6, e007328 (2021).PubMed PubMed Central Google Scholar 
  10. Assebe, L. F. et al. Cost-effectiveness of TB diagnostic technologies in ethiopia: a modelling study. Cost-effectiveness Resource Allocation: C/E. 22, 43. https://doi.org/10.1186/s12962-024-00544-1 (2024).Article PubMed Google Scholar 
  11. Parry, C. M. Sputum smear negative pulmonary tuberculosis. Trop. Doct. 23, 145–146. https://doi.org/10.1177/004947559302300402 (1993).Article ADS PubMed Google Scholar 
  12. Shaweno, D., Trauer, J. M., Denholm, J. T. & McBryde, E. S. A novel bayesian Geospatial method for estimating tuberculosis incidence reveals many missed TB cases in Ethiopia. BMC Infect. Dis. 17, 662. https://doi.org/10.1186/s12879-017-2759-0 (2017).Article PubMed PubMed Central Google Scholar 
  13. People’s Health Movement. Health Workers, Tigray.
  14. Legesse, A. Y. et al. Lived experience of healthcare providers amidst war and siege: a phenomenological study of ayder comprehensive specialized hospital of tigray, Northern Ethiopia. BMC Health Serv. Res. 292 (2024).
  15. UNICEF. Ethiopia Humanitarian Situation: Tigray Crisis Situation Report No.2. United Nations International Children’s Emergency Fund (UNICEF). https://www.unicef.org/media/92186/file/UNICEF-Ethiopia-Humanitarian-Situation-Report-No.-2-Tigray-Crisis-14-31-January-2021.pdf (2021).
  16. Adams, S. T. & Leveson, S. H. Clinical prediction rules. BMJ (Clinical Res. ed.). 344, d8312. https://doi.org/10.1136/bmj.d8312 (2012).Article Google Scholar 
  17. WHO. Mid-level Health Providers a Promising Resource To Achieve the Health Millennium Development Goals: Global Health Workforce Alliance, Report (World Health Organization, 2010). https://www.who.int/workforcealliance/knowledge/resources/Final_MLP_web_2.pdf
  18. Gebregergs, G. B., Berhe, G., Gebrehiwot, K. G. & Mulugeta, A. A qualitative study to inform the development of a decision support tool for the diagnosis of pulmonary tuberculosis in Tigray, Ethiopia. BMC Med. Informat. Dec. Making 24:338, (2024).
  19. Balcha, T. T. et al. A clinical scoring algorithm for determination of the of tuberculosis in HIV-Infected adults: A cohort study performed at Ethiopian health centers. Open. Forum Infect. Dis. 1, ofu095. https://doi.org/10.1093/ofid/ofu095 (2014).Article PubMed PubMed Central Google Scholar 
  20. Wolde, H. F., Clements, A. C. A. & Alene, K. A. Development and validation of a risk prediction model for pulmonary tuberculosis among presumptive tuberculosis cases in Ethiopia. BMJ Open. 13, e076587. https://doi.org/10.1136/bmjopen-2023-076587 (2023).Article PubMed PubMed Central Google Scholar 
  21. Akamo, J. O. & Fisseha, T. The EU’s Entry Point into the Post-Tigray War Context. (2023).
  22. Riley, R. D. et al. Minimum sample size for developing a multivariable prediction model: PART II – binary and time-to-event outcomes. Stat. Med. 38, 1276–1296. https://doi.org/10.1002/sim.7992 (2018).Article MathSciNet PubMed PubMed Central Google Scholar 
  23. Steingart, K. R. et al. Xpert® MTB/RIF assay for pulmonary tuberculosis and rifampicin resistance in adults. The Cochrane database of systematic reviews 2014, Cd009593, https://doi.org/10.1002/14651858.CD009593.pub3 (2014).
  24. Gebregergs, G. B., Berhe, G., Gebrehiwot, K. G. & Mulugeta, A. Predictors contributing to the Estimation of pulmonary tuberculosis among adults in a resource-limited setting: A systematic review of diagnostic predictions. SAGE Open. Med. 12, 20503121241243238. https://doi.org/10.1177/20503121241243238 (2024).Article PubMed PubMed Central Google Scholar 
  25. Fonti, V. & Belitser, E. Feature Selection using LASSOVU Amsterdam, Research Paper in Business Analytics,. (2017).
  26. Steyerberg, E., Bleeker, S., Moll, H. A., Grobbee, D. & Moons, K. Internal and external validation of predictive models: a simulation study of bias and precision in small samples. J. Clin. Epidemiol. 56, 441–447 (2003).PubMed Google Scholar 
  27. Vickers, A. J. & Elkin, E. B. Decision curve analysis: a novel method for evaluating prediction models. Med. Decis. Making. 26, 565–574 (2006).PubMed PubMed Central Google Scholar 
  28. Collins, G. S., Reitsma, J. B., Altman, D. G. & Moons, K. G. M. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMC Med. 13, 1. https://doi.org/10.1186/s12916-014-0241-z (2015).Article PubMed PubMed Central Google Scholar 
  29. Fletcher, R. H., Fletcher, S. W. & Fletcher, G. S. Clinical Epidemiology, the Essentials (Lippincott Williams & Wilkins, 2014).
  30. Zhu, M. et al. Times series analysis of age-specific tuberculosis at a rapid developing region in china, 2011–2016. Sci. Rep. 8, 8727. https://doi.org/10.1038/s41598-018-27024-w (2018).Article ADS PubMed PubMed Central Google Scholar 
  31. WHO. WHO Releases New Global Lists of high-burden Countries for TB, HIV-associated TB, and drug-resistant TB (World Health Organization, 2021).
  32. Risher, K. A. et al. Age patterns of HIV incidence in Eastern and Southern africa: a modelling analysis of observational population-based cohort studies. Lancet HIV. 8, e429–e439. https://doi.org/10.1016/s2352-3018(21)00069-2 (2021).Article PubMed PubMed Central Google Scholar 
  33. Lyon, S. M. & Rossman, M. D. Pulmonary tuberculosis. Microbiol. Spectr. 5 https://doi.org/10.1128/microbiolspec (2017).
  34. Baik, Y. et al. A clinical score for identifying active tuberculosis while awaiting Microbiological results: development and validation of a multivariable prediction model in sub-Saharan Africa. PLoS Med. 17, e1003420. https://doi.org/10.1371/journal.pmed.1003420 (2020).Article PubMed PubMed Central Google Scholar 
  35. Esmail, H., Dodd, P. J. & Houben, R. Tuberculosis transmission during the subclinical period: could unrelated cough play a part? Lancet Respiratory Med. 6, 244–246. https://doi.org/10.1016/s2213-2600(18)30105-x (2018).Article Google Scholar 
  36. Wang, Z., Wang, M., Wen, S., Yu, L. & Xu, X. Types and applications of cough-related questionnaires. J. Thorac. Disease. 11, 4379–4388. https://doi.org/10.21037/jtd.2019.09.62 (2019).Article Google Scholar 
  37. Zimmer, A. J. et al. Making cough count in tuberculosis care. Commun. Med. 2, 83. https://doi.org/10.1038/s43856-022-00149-w (2022).Article PubMed PubMed Central Google Scholar 
  38. Kaswaswa, K. et al. Effect of patient-delivered household contact tracing and prevention for tuberculosis: A household cluster-randomised trial in Malawi. PLoS One. 17, e0269219 (2022).PubMed PubMed Central Google Scholar 
  39. Boyles, T. H. et al. A clinical prediction score including trial of antibiotics and C-Reactive protein to improve the diagnosis of tuberculosis in ambulatory people with HIV. Open. Forum Infect. Dis. 7, ofz543. https://doi.org/10.1093/ofid/ofz543 (2020).Article PubMed PubMed Central Google Scholar 
  40. Divala, T. H. et al. Utility of broad-spectrum antibiotics for diagnosing pulmonary tuberculosis in adults: a systematic review and meta-analysis. Lancet Infect. Dis. 20, 1089–1098. https://doi.org/10.1016/s1473-3099(20)30143-2 (2020).Article PubMed PubMed Central Google Scholar 
  41. Long, R. & Cowie, R. Tuberculosis: 4. Pulmonary disease. CMAJ 160, 1344–1348 (1999).
  42. Tedla, K., Medhin, G., Berhe, G., Mulugeta, A. & Berhe, N. Factors associated with treatment initiation delay among new adult pulmonary tuberculosis patients in tigray, Northern Ethiopia. PLoS One. 15, e0235411. https://doi.org/10.1371/journal.pone.0235411 (2020).Article PubMed PubMed Central Google Scholar 
  43. Ouyang, J. et al. The development and validation of a diagnostic scoring system to differentiate pulmonary tuberculosis from non-tuberculosis pulmonary infections in HIV-infected patients with severe immune suppression. BMC Infect. Dis. 21, 863. https://doi.org/10.1186/s12879-021-06552-3 (2021).Article PubMed PubMed Central Google Scholar 
  44. Yu, Q., Guo, J. & Gong, F. Construction and validation of a diagnostic scoring system for predicting active pulmonary tuberculosis in patients with positive T-SPOT based on indicators associated with coagulation and inflammation: A retrospective Cross-Sectional study. Infect. Drug Resist. 16, 5755–5764. https://doi.org/10.2147/idr.s410923 (2023).Article PubMed PubMed Central Google Scholar 
  45. WHO. High-priority Target Product Profiles for New Tuberculosis Diagnostics: Report of a Consensus Meeting (World Health Organization, 2014).
  46. McNerney, R. & Daley, P. Towards a point-of-care test for active tuberculosis: Obstacles and opportunities. Nat. Rev. Microbiol. 9, 204–213. https://doi.org/10.1038/nrmicro2521 (2011).Article PubMed Google Scholar 
  47. Castro, C. B. A., Costa, P. A., Ruffino-Netto, A., Maciel, E. L. N. & Kritski, A. L. Assessment of a clinical score for screening suspected pulmonary tuberculosis cases. Rev. Saúde Pública. 45, 1110–1116 (2011).PubMed Google Scholar 

Download references

Acknowledgements

The authors are grateful to the Tigray National Regional State Health Bureau and all the heads of the selected health facilities for their unreserved help during data collection. We also acknowledge the NORAD Project-Mekelle University for their financial support of the study. In addition, we appreciate the willingness of the study subjects to take part in this study. Finally, we thank Mr. Hailu Abraha for proofreading and improving the manuscript’s clarity and readability.

Author information

Authors and Affiliations

  1. School of Public Health, College of Health Sciences, Mekelle University, Mekelle, EthiopiaGebremedhin Berhe Gebregergs, Gebretsadik Berhe & Afework Mulugeta
  2. Department of Internal Medicine, College of Health Sciences, Mekelle University, Mekelle, EthiopiaKibrom Gebreslasie Gebrehiwot

Contributions

GBG designed and conceptualized the study, gathered the data, carried out the analysis, and drafted the manuscript. All the authors critically revised the manuscript and approved the submitted version.

Corresponding author

Correspondence to Gebremedhin Berhe Gebregergs.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Gebregergs, G.B., Berhe, G., Gebrehiwot, K.G. et al. Development and validation of a risk prediction model for pulmonary tuberculosis in presumptive tuberculosis patients in Tigray, northern Ethiopia. Sci Rep 15, 32270 (2025). https://doi.org/10.1038/s41598-025-17959-2

Download citation