Tuberculosis Diagnostic Delays in Indonesian Primary Care Pathways

Study design and setting

This facility‑based, register‑based observational study used data from the national presumptive tuberculosis register at a primary health centre in Tasikmalaya City, West Java, Indonesia. The dataset covered 1 January to 31 December 2025 and reflected routine fields used for presumptive tuberculosis registration, diagnostic testing, and care cascade monitoring. The manuscript was prepared in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology reporting principles.

Study population

Episodes were excluded if duplicate records, missing registration dates, missing diagnostic result dates, or implausible chronological sequences were identified during data cleaning. Additional records with missing covariate information were excluded from multivariable analyses. The selection process and reasons for exclusion are presented in Figure 1. The study population comprised all people entered into the presumptive tuberculosis register during the study period. A presumptive tuberculosis episode was defined as a person with clinical or epidemiological suspicion of tuberculosis who was assigned a presumptive tuberculosis registration number. For the present analysis, episodes were eligible if at least one diagnostic test had been ordered and documented according to the national tuberculosis diagnostic algorithm. Individuals registered as presumptive tuberculosis cases but without any documented diagnostic test order were not included because the primary outcome required a measurable interval between registration and the first recorded diagnostic result. Episodes with duplicate records, missing key dates, or implausible chronological sequences were excluded during data cleaning [4]. Episodes were excluded from delay analyses if registration or diagnostic dates were missing, duplicated, or implausible.

Figure 1

Data cleaning and missing data

Data quality checks were performed before analysis. Duplicate records were identified using presumptive tuberculosis registration numbers and demographic information. Records with missing registration dates, missing diagnostic result dates, or implausible chronological sequences (for example, diagnostic results recorded before registration) were excluded from delay analyses. Missing covariate information was handled using complete‑case analysis for multivariable models. The numbers of exclusions at each stage are presented in Figure 1.

Variables and outcomes

Variables included age, sex, residential area, referral source, anatomical disease site, previous tuberculosis treatment status, documented diabetes, first diagnostic test, diagnostic result, final diagnosis, and treatment initiation. Diagnostic tests recorded in the register included sputum smear microscopy, Xpert MTB/RIF (GeneXpert), chest radiography, mycobacterial culture, histopathology, and other investigations performed according to the Indonesian National Tuberculosis Diagnostic Algorithm. The first diagnostic test variable referred to the earliest documented investigation requested following presumptive tuberculosis registration.

Age was analysed continuously and in life‑course groups: 0–14, 15–24, 25–44, 45–64, and 65 years or older. The primary outcome was diagnostic delay, defined as the number of days from registration as a presumptive tuberculosis case to the first recorded diagnostic result. When more than one diagnostic result was recorded, the earliest valid result was used. Same‑day results were assigned a delay of zero days. Prolonged delay was defined as more than 14 days, with more than 30 days assessed in sensitivity analyses. The >14‑day threshold was selected because it has been commonly used in studies evaluating tuberculosis diagnostic delays and represents a pragmatic operational benchmark for identifying potentially avoidable delays within the diagnostic pathway. A >30‑day threshold was examined in sensitivity analyses to assess more substantial delays and facilitate comparison with previous tuberculosis delay literature. Secondary outcomes were receipt of a final tuberculosis diagnosis among presumptive episodes and documented treatment initiation among diagnosed patients. This measure captures the interval after entry into the formal diagnostic system and may therefore be interpreted as a facility‑level diagnostic turnaround indicator.

Statistical analysis

Characteristics were summarised using medians and interquartile ranges or frequencies and percentages. Group differences were assessed using the Mann‑Whitney U, Kruskal‑Wallis, chi‑square, or Fisher exact tests, as appropriate. Negative binomial regression estimated incidence rate ratios for diagnostic delay because the outcome represented count data (days) with a right‑skewed distribution. Overdispersion was assessed by comparing the variance and mean of the delay variable and was observed to be present; therefore, negative binomial regression was preferred over Poisson regression. Logistic regression estimated adjusted odds ratios for prolonged delay, final diagnosis, and treatment initiation. Models included age group, sex, anatomical disease site, referral source, diabetes, and residence. Sensitivity analyses used the more than 30‑day delay threshold and excluded records with missing key covariates. Statistical significance was assessed using two‑sided tests with alpha = 0.05. Statistical analyses were conducted using Stata version 18.0 (StataCorp LLC, College Station, TX, USA).

Ethics and consent

The study was approved by the Health Research Ethics Committee, Faculty of Public Health, Universitas Muhammadiyah Jakarta (approval number: 10.166.C/KEPK‑FKMUMJ/I/2026; approval date: 30 January 2026). Permission to access and analyse register data was obtained from the Tasikmalaya City Health Office and the participating primary health centre. Direct identifiers were removed before analysis. Individual informed consent was waived because the study used retrospective, de‑identified secondary data consistent with Indonesian health data governance requirements [17, 18].

Principal findings

This register‑based study found that diagnostic results were generally recorded within a short period after presumptive tuberculosis registration in primary care, but nearly one in five episodes still experienced a delay beyond 14 days. Older age, extrapulmonary disease, and referral from another health facility were the clearest predictors of longer post‑registration delay. These findings suggest that even after a person enters the formal tuberculosis diagnostic pathway, specific patient groups and referral pathways remain vulnerable to avoidable diagnostic friction.

Interpretation in relation to prior evidence

The median post‑registration delay observed in this study was shorter than estimates from studies measuring time from symptom onset or first care seeking. Systematic reviews and Indonesian pathway studies have reported much longer total diagnostic intervals, often reflecting the cumulative effects of symptom appraisal, repeated provider visits, cost, stigma, and navigation across public and private services [7, 8, 19]. The present study, therefore, should not be interpreted as measuring total patient delay. Rather, it isolates a later operational segment: the time from formal presumptive tuberculosis registration in primary care to the availability of the first recorded diagnostic result.

Equity implications for older adults and extrapulmonary tuberculosis

Older adults had longer delays, which is clinically plausible and programmatically important. Tuberculosis in older people may present with non‑specific symptoms, coexisting chronic diseases, frailty, or difficulty producing sputum, all of which can reduce diagnostic suspicion or slow specimen collection and follow‑up [20]. Primary health centre teams could therefore flag older presumptive patients for active tracking when results are not available within two weeks. Extrapulmonary disease was also associated with longer delay. The finding that extrapulmonary disease was associated with lower odds of receiving a final tuberculosis diagnosis warrants particular attention. Extrapulmonary tuberculosis frequently requires imaging, histopathology, or specialist assessment that may not be available within primary care facilities. Some patients may therefore be referred for further evaluation without complete documentation returning to the primary care register. This could contribute both to genuine diagnostic complexity and to under‑ascertainment of confirmed diagnoses within routine primary care records. This aligns with evidence that extrapulmonary tuberculosis is more difficult to confirm because symptoms are site‑specific, microbiological yield may be low, and diagnosis may require imaging, tissue sampling, or referral to higher‑level facilities [21]. Clear escalation pathways for suspected extrapulmonary tuberculosis are needed so that referral does not become an unmonitored waiting period.

Referral pathways and continuity of care

Referral source mattered. Patients entering from other health facilities had longer delays than self‑presenting patients, and private‑provider referrals showed elevated but less precise associations. Indonesian studies show that people with tuberculosis commonly seek care from private clinics, hospitals, pharmacies, or informal providers before reaching public tuberculosis services [7, 22]. These findings are highly relevant to public‑private mix strategies because referral alone is not sufficient unless diagnostic requests, specimen movement, result return, and treatment linkage are actively closed [23, 24]. Standardised referral forms, date‑stamped specimen requests, interoperable reporting, and feedback mechanisms may reduce losses between providers.

An additional continuity‑of‑care concern relates to treatment initiation after diagnosis. In this study, 18 of 176 diagnosed patients (10.2%) had no documented treatment initiation. Although the available data could not determine whether these individuals failed to start treatment or whether treatment initiation was not recorded, this finding may represent potential initial loss to follow‑up. Patients who do not promptly initiate treatment remain at risk of disease progression, ongoing transmission, and disengagement from tuberculosis care services. Strengthening linkage‑to‑care mechanisms and treatment documentation systems may help reduce these losses within the tuberculosis care cascade.

Global health relevance

From a global health perspective, the study highlights a practical equity‑monitoring opportunity within routine primary care systems. Many high‑burden countries rely on primary care registers and referral networks to close tuberculosis detection gaps. Register‑based indicators can be used not only to count presumptive cases and confirmed diagnoses, but also to detect where specific groups experience slower diagnostic progression. This is especially relevant in lower‑ and middle‑income countries, where diagnostic decentralisation, molecular testing access, and public‑private coordination remain uneven [25–28]. Accordingly, our findings should not be interpreted as total tuberculosis diagnostic delay but rather as delay occurring after registration within the formal diagnostic pathway.

Implications of diagnostic delays

Although the median diagnostic interval was relatively short, prolonged delays remain clinically and programmatically important. Delays in obtaining diagnostic results may postpone treatment initiation among patients ultimately diagnosed with tuberculosis, increase opportunities for initial loss to follow‑up, and prolong periods during which infectious individuals remain untreated within households and communities. Delayed diagnostic results may also postpone the identification and management of alternative conditions among patients who do not have tuberculosis, potentially contributing to continued morbidity and repeated healthcare utilisation. In addition, prolonged waiting times may reduce patient confidence in health services and increase the likelihood of disengagement from care. These consequences highlight why monitoring diagnostic timeliness remains an important component of tuberculosis care cascade improvement, even after individuals have entered the formal diagnostic system.

Strengths and limitations

The main strength of this study is its use of routine presumptive tuberculosis register data, which captures an operational point that is often missed in studies restricted to confirmed patients. An additional limitation relates to the use of routine register data. The primary outcome measured the interval from presumptive tuberculosis registration to the first recorded diagnostic result rather than the exact time at which a diagnostic result became available. Consequently, the observed interval may reflect delays in documentation, reporting, or data entry in addition to clinical or health‑system diagnostic processes. The analysis also examined multiple care cascade outcomes, including diagnostic delay, final diagnosis, and treatment initiation. However, several limitations should be considered. First, the findings cannot be interpreted as comprehensive programme performance because the study was conducted in one primary health centre. Second, the register‑based design measured delay from registration to diagnostic result and did not capture symptom onset, pre‑registration care seeking, cost, stigma, distance, or provider switching. Third, routine registers may contain missing or inconsistently updated referral, comorbidity, and treatment fields. In addition, HIV status was incompletely recorded in the routine register and could not be reliably incorporated into multivariable analyses. Consequently, potential confounding related to HIV infection could not be fully assessed. Finally, because the study was conducted in a single primary health centre, the findings should be interpreted primarily as locally relevant operational evidence rather than as representative of Indonesian primary care tuberculosis services nationally. In addition, the register did not consistently capture the specific diagnostic modality associated with each recorded result. Consequently, it was not possible to determine which diagnostic tests contributed to same‑day (0‑day) results or to evaluate the extent to which variation in diagnostic turnaround time reflected differences in testing technologies. Because diagnostic modalities such as Xpert MTB/RIF, smear microscopy, culture, and other investigations have substantially different turnaround times, unmeasured variation related to diagnostic test type may have influenced the observed delay estimates.