Data for our nationwide, population-based, cohort study were obtained from the inpatient care records and registration files from the NHIRD. The National Health Insurance (NHI) program was implemented in 1995 and provides healthcare coverage to 99% of the population (being more than 23 million people). The NHI Administration randomly reviews the records of ambulatory care visits and in-patient claims periodically so as to verify the accuracy of the diagnoses (National Health Insurance Administration National Health Insurance Regulations). The details of the program have been documented in previous studies (Chang et al., 2018; Chao et al., 2018). It should be noted that some studies have demonstrated the accuracy and validity of several diagnoses in the NHIRD, including diabetes mellitus (DM), cancer, myocardial infarction, and the central nervous system diseases, such as Tourette syndrome, and stroke.

This study involved a cohort design. Using the NHIRD, we selected adult patients aged >50 years who had been diagnosed with AD (331.0) based on the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) codes between 2000 and 2015, and confirmed these patients' diagnoses by linking the records of at least three visits for AD in consecutive years, or at least one hospitalization for AD, during the study period. All diagnoses of AD were made by board-certified psychiatrists or neurologist in Taiwan. The date of the AD diagnosis was at least 1 year after the CPn pneumonia diagnosis. We identified patients with CPn pneumonia using the ICD-9-CM code 483.1. We confirmed these patients' diagnoses by linking the records of at least one hospitalization for CPn pneumonia, in the study period, and the reason why we decided on the enrolled exposures of CPn as pneumonia due to CPn, was because one previous population-based study had confirmed the accuracy of the diagnosis of pneumonia from the NHIRD as 98.3% (Su et al., 2014). For each patient with Cpn pneumonia included in our study, three controls were selected via 1:3 matching by age, sex, and the number of medical follow-ups (N = 4971) in the NHIRD. All insurance claims were scrutinized by medical reimbursement specialists, and peer reviews were undertaken according to the standard and clinical diagnostic criteria.

The covariates included sex, age groups (50–64, ≥65 years), marital status, education (<12 years,≥12 years), seasons, geographical area of residence (north, center, south, and east of Taiwan), urbanization level of residence (levels 1–4), and monthly income (in New Taiwan Dollars [NT$]; <18,000, 18,000–34,999, ≥35,000). The urbanization level of residence was defined according to the population and various indicators of the level of development. Level 1 was defined as a population of >1,250,000, and a specific designation as political, economic, cultural, and metropolitan development. Level 2 was defined as a population between 500,000 and 1,249,999, and as playing a key role in the politics, economy, and culture. Urbanization levels 3 and 4 were defined as a population between 149,999 and 499,999, and <149,999, respectively (Chang et al., 2014).

Data on the usage of macrolides and fluoroquinolones antibiotics were acquired from the Longitudinal Health Insurance Database (LHID), a sub-database of the NHIRD. The data of the defined daily dosage (DDD) were obtained from the WHO Collaborating Centre for Drug Statistics Methodology (https://www.whocc.no/), and the duration of the usage of antibiotics was calculated by dividing the cumulative dosages by the DDD of the antibiotics.

The comorbidities, including DM (ICD-9-CM 250), hypertension (ICD-9-CM 401.1, 401.9, 402.10, 402.90, 404.10, 404.90, 405.1, and 405.9), hyperlipidemia (ICD-9-CM 272), coronary artery disease (CAD, ICD-9-CM code 410–414), obesity (ICD-9-CM 278), all cancers (ICD-9-CM 140–208), and other chlamydia infections (ICD-9-CM codes: 077.98, ICD-9-CM 078.88, 079.88, 079.98, 099.41, and 099.5) and other pneumonia (ICD-9-CM codes: 480–486, except 483.1). These comorbidities were included with the references from previous studies using health databases (Wright et al., 2015; Gottesman et al., 2017).

The Charlson comorbidity index (CCI) was employed to execute the assessment of the aforementioned comorbidities. In the CCI, ICD-9-CM codes are used as the basis for the establishment of categories for comorbidities, and each of the established categories is scored (van den Berg et al., 2013, 2014; Wong et al., 2014); to derive a single comorbidity score, all CCI scores are combined, with 0 indicating no comorbidities and higher scores (1, 2, 3, ≥4) indicating higher comorbidity burdens (Needham et al., 2005).

Only patients diagnosed with CPn pneumonia of more than 1 year prior to the index date were considered. CPn pneumonia was identified from the NHIRD by using the corresponding ICD-9 code (ICD-9-CM code 483.1). All the covariates as aforementioned were included.

Categorical variables, which were presented as percentages, were compared using the χ² tests and the Fisher's exact test. Continuous variables, which were presented as the mean and SD, were compared using the Student's t-tests. The primary goal of this study was to determine as to whether a patient's clinical characteristics, such as CPn pneumonia, were associated with AD. Associations between those outcomes and clinical characteristics were investigated using the Fine and Gray's survival analysis in a generalized estimating equation (GEE) model. The regression results are presented as adjusted HRs with corresponding 95% CIs. The threshold for statistical significance was p < 0.05. All data analyses were conducted using the SPSS V.22 (SPSS).

Supplementary Figure 1 is a flowchart of the patient enrollment procedure. From the NHIRD, we identified 1,657 patients who received a CPn pneumonia diagnosis during our defined study period (2000–2015); these patients were matched 1:3 with patients without CPn (N = 4,971) according to age, sex, number of visits to medical facilities, and comorbidities.

In total, 1,657 patients were diagnosed with Cpn pneumonia during the study period. A total of 182 enrollee were identified with AD during the follow-up period.

Table 1 shows the sex, age, marital status and comorbidities of the patients with Cpn pneumonia. When compared with controls, the patients with Cpn pneumonia tended to have higher rate of hypertension, hyperlipidemia, other pneumonia, anxiety, sleep disorder and CCI scores.

Patients with CPn pneumonia also tended to be living in northern and middle Taiwan and residing more in the regions of urbanization levels 1 and 2. There were no differences in the distribution of sex, age, marital status, and education between these two groups.

In our applied Fine & Gray's competing risk model, compared with the control group, we determined patients with CPn pneumonia to have a higher risk of AD (adjusted hazard ratio [HR] = 1.599, 95% CI = 1.284–1.971, p < 0.001; Table 2) after we adjusted for urbanization level/geographic region, sex, marital status, education, comorbidities (including CCI scores), age, antibiotic use, and insurance premium.

In addition, we observed patients with hyperlipidemia, CAD, anxiety, sleep disorder, and alcohol use disorder to be at a relatively high risk of AD. Furthermore, we noted those receiving care from a medical center or regional hospital to exhibit a relatively high risk of AD. Conversely, patients who sought medical care in Autumn and with cancer revealed a reduced risk of AD. Although hypertension exhibited a significant difference in the patients with CPn pneumonia group at baseline (see in Table 1), it did not significantly affect developing AD by the logistic regression model (p = 0.087; Table 2).

Table 3 shows the usage of antibiotics in the risk of AD in patients with CPn pneumonia, and that the usage of macrolides (≥15 days) and fluoroquinolones (≥15 days), were associated with a decreased risk of dementia.

The cumulative incidences of AD were 931.01 per 10⁵ person-years and 764.22 per 10⁵ person-years, in the study cohort and comparison cohort group, respectively (Supplementary Table 1). The difference between the two groups was significant (log-rank test, p < 0.001; Figure 1). After 8 years of tracking, the cumulative incidence of AD in patients with CPn pneumonia and comparison group was significant (log-rank test, p < 0.037; Figure 1).

By employing a nationwide population database, our executed study is the first to demonstrate an association between CPn pneumonia and AD development in Taiwan. Patients with CPn pneumonia exhibited a nearly 1.6-fold increased risk of AD (overall adjusted HR = 1.599; 95% CI = 1.284–1.971, p < 0.001). Additionally, hyperlipidemia, CAD, anxiety, sleep disorder, and alcohol use disorder exhibited associations with a relatively high risk of AD. Patients who visited medical centers or regional hospitals to seek medical care were also noted to be at relatively high risk. Although the percentage of hypertension in CPn exposed cohort was higher than that in the unexposed cohort (p < 0.001), in Fine & Gray's survival analysis, there were no significant differences for hypertension between the two cohorts (p = 0.087). However, patients who sought medical care in Autumn and with cancer revealed a reduced risk of AD. Further studies are needed to clarify why patients with these comorbidities were associated with a decreased risk. Moreover, our study also indicated that patients who were treated with macrolides (≥15 days) and fluoroquinolones (≥15 days) could attenuate the risk of AD.

Some previous studies have found that CPn was associated with AD (Balin et al., 1998; Gerard et al., 2006; Paradowski et al., 2007), but other studies had failed to detect such an association (Taylor et al., 2002; Wozniak et al., 2003; Yamamoto et al., 2005; Hammond et al., 2010). In one meta-analysis—the procedures of which entailed pooling cases and controls in previous studies—chlamydial infection was noted to exhibit an association with a 5-fold rise in the AD occurrence (OR: 5.66; 95% CI = 1.83–17.51, p < 0.001) (Maheshwari and Eslick, 2015). In the present study, because the pneumonia diagnosis as recorded in the NHIRD was determined to have 98.3% accuracy (Su et al., 2014), we restricted patients with CPn infection to those with CPn pneumonia to bolster the credibility of the association between AD and CPn.

The first formal characterization of AD was made in 1907; nevertheless, its corresponding etiology remains poorly understood. AD is commonly regarded as a neurodegenerative disease and is ascribed to neuronal damage and death. Neuropathology in AD is characterized by neuropil threads, neurofibrillary tangles, neurotic senile plaques, and typically amyloid deposits around the cerebral vasculature. Neuroinflammatory processes, oxidative stress, and vascular factors are the three main contributors to AD pathogenesis (Ashraf et al., 2019).

Many studies have confirmed the relationship between neuroinflammation and AD, despite the mechanistic links requiring further investigation. Herpes simplex type 1, human herpesvirus 6, Borrelia burgdorferi, and Helicobacter pylori are among the myriad of pathogens that may be involved in AD (Mawanda and Wallace, 2013). Neuroinflammation in the brain may contribute to AD pathogenesis. Among various pathogens, researchers have proposed CPn may cause chronic neuroinflammation and be implicated in AD. CPn infection of endothelial cells—infection at the vascular level—could engender increased monocyte migration and advance inflammation in the CNS (MacIntyre et al., 2003). Moreover, in THP1 human monocytes, CPn infection stimulates an innate immune response, and thus, such infection may have a role in inflammation initiation in cases of sporadic/late-onset Alzheimer disease (Lim et al., 2014). Some animal studies have determined that CPn is related to beta-amyloid (Aβ) 1–42 immunoreactive deposits in rodent brain tissues (Little et al., 2004; Boelen et al., 2007). A recent in vitro research claimed that CPn infection also altered the expression and activity of amyloid precursor protein and promoted amyloid formation (Al-Atrache et al., 2019). These findings provided evidence for a link between CPn and AD pathology.

According to the antimicrobial protection hypothesis (Moir et al., 2018), Aβ oligomerization is not inherently pathological; Aβ deposition represents an early innate immune reaction to an immunochallenge, whether genuine or mistakenly perceived. Aβ fibrillization inactivates neuroinflammatory pathways to defend against infection and clear the deposition of Aβ or pathogens. In cases of AD, sustained inflammation and neurodegeneration result from chronic activation of the pathway. It could be explained the cumulative incidence of AD in patients with CPn pneumonia exhibited significant after 8 years of tracking in this study.

In addition, the apolipoprotein E4 (APOE) genotype may be involved in the CPn–AD association (Wehr et al., 2006). Scholars have determined the APOE gene to be involved in cerebral Aβ clearance, with CPn influencing neuronal damage repair ability (Masters et al., 2015; Woods et al., 2020). Hence, another possible mechanism underlying CPn toxicity in AD may be the interaction of the APOE genotype with CPn.

As revealed in the literature, antibiotics against CPn were noted to decrease the deposition of amyloids, oxidative stress, and inflammation in AD models (Yulug et al., 2018; Balducci and Forloni, 2019); these antibiotics possess the capacity to traverse the blood–brain barrier. In this study, we noted macrolide use (≥15 days) and fluoroquinolone use (≥15 days) to exhibit an association with a decrease in dementia risk. Compared with macrolides, fluoroquinolones are much more lipophilic and smaller and are hence more efficacious when administered in CNS. Macrolides cannot cross the blood–brain barrier, but they are usually the first-line therapy for CPn infection; no evidence has suggested macrolides are neurotoxic or affect AD risk, and it is presumed such drugs (i.e., azithromycin, clarithromycin) do not exacerbate AD (Woods et al., 2020). The role of antibiotics against CPn and the risk of AD requires further investigation.

Pneumonia caused by CPn is typically mild, and most patients recover without complications. Our study indicated that the appropriate use of antibiotics with macrolides and fluoroquinolones (both duration≥15 days) could decrease the risk of developing AD. In one clinical practice guideline, continue antibiotic therapy is strongly recommended in patients with CPn infection until they achieve clinical stability including resolution of vital sign abnormalities, appetite, and normal mentation (Metlay et al., 2019).

CPn is a type of community-acquired pneumonia and is transmitted person-to-person through inhalation of respiratory droplets or contact with droplets on surfaces followed by contact (i.e., touching) with the mouth or nose. CPn reinfection among older adults is common, and outbreaks have been reported in settings of close contact and crowding, such as nursing homes, schools, prisons, and military barracks (Burillo and Bouza, 2010). This transmissibility may explain why we noted an association between a high AD risk undergoing medical care at regional hospitals or medical centers.

The NHIRD recorded inpatient care, ambulatory care, dental care, and prescription drugs availed by the insured and their date of birth. However, pursuant to the Personal Information Protection Act, individual identifiers are encrypted before releasing for research.

Therefore, information such as weakness severity, laboratory parameters, neurological symptom severity, additional examination findings (e.g., electrophysiological testing), or rehabilitation availability could not be assessed in our executed study due to the lack of such data in the NHIRD.

Besides, we could not include data on psychosocial, environmental, and genetic factors in our analyses due to the same reason. However, despite these limitations, our derived data are highly likely to be valid and representative due to the NHIRD containing data covering all hospitals within Taiwan and over 99% of the population for the relevant 15-year period.