**Background:** Healthcare-associated infections (HAIs) affect approximately 10% of patients in developed countries and 25% in developing countries, with risk 2–20 fold higher in resource-limited settings. In Bangladesh, HAI rates may exceed 30% in some facilities. Community clinics (CCs) are the most basic primary health facilities serving rural populations, yet no prior studies had examined infection prevention (IP) practices among community healthcare providers (CHCPs) in Bangladesh. This study aimed to assess IP knowledge and practices and identify associated factors among CHCPs in the poorest district of Bangladesh.
**Methods:** A cross-sectional study was conducted among 128 CHCPs from 128 CCs across five Upazilas (Kurigram Sadar, Nageshwari, Bhurungamari, Phulbari, Ulipur) in Kurigram district—the poorest district in Bangladesh with a 70.8% poverty rate. A stratified random sampling technique was used. Data were collected November–December 2019 via face-to-face survey using a pre-tested semi-structured questionnaire. The questionnaire covered socio-demographics, clinic resources, IP knowledge (10 items), and self-reported IP practices (13 items). Knowledge and practice scores were dichotomized at the mean into adequate/inadequate and good/poor, respectively. Chi-square, Fisher's exact tests, binary and multiple logistic regression were used. Reliability coefficients (Cronbach's Alpha) were 0.768 for knowledge and 0.753 for practice items.
**Key Results:** The mean age of participants was 32.6 ± 3.7 years; 57.8% were male; 96.1% were Muslim; 43.8% had bachelor's, 33.6% master's, and 22.7% higher secondary education. Only 7% of CCs had IP guidelines; 88.3% had hand washing facilities; 85.9% had soap always available; only 13.3% had gloves and 7% had masks always available. Only 37.5% of CHCPs had adequate IP knowledge (mean score 5.17 ± 1.38 out of 9). Notably, only 42.2% knew gloves do not provide complete protection; 47.7% recognized that wearing gloves does not replace hand washing; 55.5% believed alcohol-based antiseptic is as effective as soap and water when hands are not visibly dirty; only 21.9% knew how to prepare 0.5% chlorine solution; and 50% knew safety boxes should be sealed when three-quarters full. Only 39.1% reported good IP practices. Specifically, 57.9% washed hands before patient care; 86.7% washed after patient care; 57.8% always used aprons, 25% gloves, and 7% masks when splashes were likely; 7.8% used IP guidelines; 55.5% recapped needles; 28.9% had a history of blood/body fluid/needle stick injury, of whom only 24.3% underwent PEP; 86.7% provided health education on HAIs; 25.8% were vaccinated against common pathogens; 96.1% placed sharps in safety boxes; and 55.5% disposed of boxes when three-quarters full. Good IP practices were significantly associated with education level (χ² = 8.541, p = 0.014), hand washing facility availability (χ² = 4.725, p = 0.030), soap availability (χ² = 4.413, p = 0.036), and IP knowledge (χ² = 9.531, p = 0.002). In the adjusted model, having a master's degree (AOR = 4.92, 95% CI: 1.41–17.23, p = 0.013) and adequate IP knowledge (AOR = 2.89, 95% CI: 1.26–6.63, p = 0.012) were significant independent predictors of good IP practices.
**Clinical Implications:** The majority of CHCPs in rural Bangladesh have inadequate IP knowledge and poor IP practices, with only about two-fifths meeting acceptable thresholds. The strong association between IP knowledge and practice suggests that targeted educational interventions could yield meaningful improvements. The findings also underscore that basic infrastructure—hand washing facilities and soap—is foundational to IP compliance. The very low rates of glove (13.3%) and mask (7%) availability, and the fact that only 7% of clinics had IP guidelines, point to systemic resource gaps. The low vaccination rate (25.8%) and low PEP uptake after exposures (24.3%) represent critical patient and provider safety concerns. The authors recommend implementing effective IP training programs, ensuring continuous supply of basic IP resources, and establishing monitoring and supervision systems. Limitations include self-reported data (susceptible to recall and social desirability bias), cross-sectional design (no causality), small sample size, and restriction to one district limiting generalizability.