Artificial intelligence readiness in Pakistan's medical and dental education: training-phase decline, a knowledge-practice paradox, and the role of digital determinants of health.
AI use among medical and dental students in Pakistan is lower in later training years, with significant barriers including lack of training and technical access.
Where it sits
this study against the rest of the kisspeptin (kp-10) corpusSummary and findings
This study assessed AI knowledge, attitudes, and use among 501 medical and dental students in Khyber Pakhtunkhwa, Pakistan. Mean scores indicated low knowledge (4.23 out of 7), favorable attitudes (32.72 out of 50), and moderate practice (15.85 out of 35) regarding AI. The study identified predictors of low AI use, including year of study, gender, and specialty.
Abstract
<h4>Background</h4>We assessed AI knowledge, attitudes, and use among medical and dental students in Pakistan, a lower-middle-income country (LMIC), and identified predictors of low AI use within Digital Determinants of Health (DDoH) and Professional Identity Formation (PIF) frameworks.<h4>Methods</h4>A cross-sectional survey of 501 medical and dental students in Khyber Pakhtunkhwa assessed knowledge, attitudes, practices, and barriers related to AI. Analyses used ANOVA and logistic regression for low AI use (score ≤ 15; 7-35 scale, higher = greater use), with sensitivity models adjusted for age, knowledge, attitude scores, and selected barriers.<h4>Results</h4>Mean (SD) scores were 4.23 (1.71) for knowledge (0-7), 32.72 (6.64) for attitude (10-50), and 15.85 (6.32) for practice (7-35); higher scores indicated greater knowledge, more favorable attitudes, and greater AI use. Practice differed across years [F(4,496) = 4.025, <i>p</i> = 0.003, η²p = 0.031], with Year 2 scoring higher than Year 4 and the final year. This exploratory model (AUC = 0.618) identified only associations. Fourth-year (OR = 2.02, 95% CI: 1.11-3.70; <i>p</i> = 0.022), female (OR = 1.51, 95% CI: 1.05-2.17; <i>p</i> = 0.025), and dental students (OR = 1.58, 95% CI: 1.01-2.50; <i>p</i> = 0.046) had higher odds of low AI use. The Year 4 association strengthened in sensitivity analyses (OR = 3.83, 95% CI: 1.66-8.85; <i>p</i> = 0.002), and higher knowledge paradoxically predicted lower use (OR = 1.52, 95% CI: 1.34-1.72; <i>p</i> < 0.001). The most significant barriers were a lack of training (71.7%) and limited technical access (69.7%).<h4>Conclusions</h4>AI use was lower in later training years and differed by gender and specialty. The inverse knowledge-practice association contradicts standard adoption models and warrants longitudinal investigation; one PIF-based hypothesis is that higher-knowledge students exercise professional restraint, although cross-sectional data cannot confirm it. The clinical-phase decline implicates ward-based norms as independent barriers. Competency-based training with structured access provisions and faculty development, particularly in clinical phase education, is needed.
Background
This paper addresses the knowledge and use of artificial intelligence (AI) among medical and dental students in Pakistan, a lower-middle-income country. Prior studies have indicated varying levels of technology adoption in medical education, but little is known about AI specifically. Understanding the barriers to AI use is crucial for improving educational outcomes and preparing future healthcare professionals.
Methods
A cross-sectional survey was conducted with 501 medical and dental students in Khyber Pakhtunkhwa. The study assessed knowledge, attitudes, practices, and barriers related to AI using a 7-point scale for knowledge, a 50-point scale for attitudes, and a 35-point scale for practice. Analyses included ANOVA and logistic regression to identify predictors of low AI use.
Results
The mean knowledge score was 4.23 (SD 1.71), the mean attitude score was 32.72 (SD 6.64), and the mean practice score was 15.85 (SD 6.32). There was a statistically significant difference in practice scores across years (F(4,496) = 4.025, p = 0.003, η²p = 0.031). Fourth-year students had an OR of 2.02 for low AI use (95% CI: 1.11-3.70, p = 0.022).
Interpretation
The findings suggest that AI use declines in later training years, which may contradict trends seen in other technology adoptions. The small effect sizes and the cross-sectional nature of the study limit the ability to draw firm conclusions. The inverse relationship between knowledge and practice raises questions about how students apply their knowledge in clinical settings.
Key findings
- Mean knowledge score was 4.23 (SD 1.71) out of 7.
- Mean attitude score was 32.72 (SD 6.64) out of 50.
- Mean practice score was 15.85 (SD 6.32) out of 35.
- Fourth-year students had an odds ratio (OR) of 2.02 for low AI use, p=0.022.
- 71.7% reported a lack of training as a significant barrier.
- Higher knowledge paradoxically predicted lower AI use with an OR of 1.52, p<0.001.
Limitations
- cross-sectional study design limits causal inference
- self-reported data may introduce bias
- small sample size may not be representative
- findings specific to one region may not generalize